Software:List of large language models

From HandWiki
Short description: none
This is a dynamic list and may never be able to satisfy particular standards for completeness. You can help by expanding it with reliably sourced entries.

A large language model (LLM) is a type of machine learning model designed for natural language processing tasks such as language generation. LLMs are language models with many parameters, and are trained with self-supervised learning on a vast amount of text.

List

For the training cost column, 1 petaFLOP-day equals 1 petaFLOP/sec × 1 day, or 8.64×1019 FLOP (floating point operations). Only the cost of the largest model is shown. The number of parameters is measured in billions,[lower-alpha 1] and the training cost is measured in petaFLOP-days.

2018

Name Release date[lower-alpha 2] Developer Template:No. of params Corpus size Training cost License[lower-alpha 3] Notes
GPT-1 2018-06-11 OpenAI 0.117 0.117B


Unknown 1[1] MIT[2]
First GPT model, decoder-only transformer. Trained for 30 days on 8 P600 GPUs.[3]
BERT 2018-10 Google 0.340 0.340B

[4]

3300000000 3.3B 
words[4]
9 9

[5]

Apache 2.0[6]

2019

Name Release date[lower-alpha 2] Developer Template:No. of params Corpus size Training cost License[lower-alpha 3] Notes
T5 2019-10 Google 11 11B

[7]

34B tokens[7] Unknown Apache 2.0[8]
Base model for Google projects like Imagen.[9]
XLNet 2019-06 Google 0.340 0.340B

[10]

3300000000 33B 
words
330 Apache 2.0[11]
An alternative to BERT; designed as encoder-only. Trained on 512 TPU v3 chips for 5.5 days.[12]
GPT-2 2019-02 OpenAI 1.5 1.5B

[13]

40GB[14] (~10000000000 10B 
tokens)[15]
28[16] MIT[17]
Trained on 32 TPUv3 chips for 1 week.[16]

2020

Name Release date[lower-alpha 2] Developer Template:No. of params Corpus size Training cost License[lower-alpha 3] Notes
GPT-3 2020-05 OpenAI 175 175B

[18]

300000000000 300B 
tokens[15]
3640[19] Proprietary
A fine-tuned variant of GPT-3, termed GPT-3.5, was made available to the public through ChatGPT in 2022.[20]

2021

Name Release date[lower-alpha 2] Developer Template:No. of params Corpus size Training cost License[lower-alpha 3] Notes
GPT-Neo 2021-03 EleutherAI 2.7 2.7B

[21]

825 GiB[22] Unknown MIT[23]
The first of a series of free GPT-3 alternatives released by EleutherAI. GPT-Neo outperformed an equivalent-size GPT-3 model on some benchmarks, but was significantly worse than the largest GPT-3.[23]
HyperCLOVA 2021-05[24] Naver 204 204B

[25]
82B[26]

560B Tokens[25] Unknown Unreleased
Its technical report says the model is the "first discovery on near 100B-scale non-English LM" by "constructing a large Korean-centric corpus of 560B tokens."[26]
GPT-J 2021-06 EleutherAI 6 6B

[27]

825 GiB[22] 200[28] Apache 2.0
Megatron-Turing NLG 2021-10[29] Microsoft and Nvidia 530 530B

[30]

338600000000 338.6B 
tokens[30]
38000[31] Unreleased
Trained for 3 months on over 2000 A100 GPUs on the NVIDIA Selene Supercomputer, for over 3 million GPU-hours.[31]
Ernie 3.0 Titan 2021-12 Baidu 260 260B

[32]

4TB Unknown Proprietary
Claude[33] 2021-12 Anthropic 52 52B

[34]

400000000000 400B 
tokens[34]
Unknown
Fine-tuned for desirable behavior in conversations.[35]
GLaM (Generalist Language Model) 2021-12 Google 1162 1162B

[36]

1600000000000 1.6T 
tokens[36]
5600[36]
Gopher 2021-12 Google DeepMind 280 280B

[37]

300000000000 300B 
tokens[38]
5833[39]

2022

Name Release date[lower-alpha 2] Developer Template:No. of params Corpus size Training cost License[lower-alpha 3] Notes
LaMDA (Language Models for Dialog Applications) 2022-01 Google 137 137B

[40]

1.56T words,[40] 168000000000 168B 
tokens[38]
4110[41] Proprietary
GPT-NeoX 2022-02 EleutherAI 20 20B

[42]

825 GiB[22] 740[28] Apache 2.0
Chinchilla scaling 2022-03 Google DeepMind 70 70B

[43]

1400000000000 1.4T 
tokens[43][38]
6805[39] Proprietary
PaLM (Pathways Language Model) 2022-04 Google 540 540B

[44]

768000000000 768B 
tokens[43]
29250 29,250

[39]

Trained for ~60 days on ~6000 TPU v4 chips.[39]
OPT (Open Pretrained Transformer) 2022-05 Meta 175 175B

[45]

180000000000 180B 
tokens[46]
310[28] Non-commercial research
GPT-3 architecture with some adaptations from Megatron. The training logbook written by the team was published.[47] The smaller models including 66B are publicly available, while the 175B model is available on request.
YaLM 100B 2022-06 Yandex 100 100B

[48]

1.7TB[48] Unknown Apache 2.0
Minerva 2022-06 Google 540 540B

[49]

38.5B tokens from webpages filtered for math content and from arXiv[49] Proprietary
For solving "mathematical and scientific questions using step-by-step reasoning".[50]
BLOOM 2022-07 Large collaboration led by Hugging Face 175 175B

[51]

350000000000 350B 
tokens (1.6TB)[52]
Responsible AI
Galactica 2022-11 Meta 120 120B


350000000000 106B 
tokens[53]
CC-BY-NC-4.0
AlexaTM (Teacher Models) 2022-11 Amazon 20 20B

[54]

1300000000000 1.3T

[55]

Proprietary[56]

2023

Name Release date[lower-alpha 2] Developer Template:No. of params Corpus size Training cost License[lower-alpha 3] Notes
Llama 2023-02 Meta AI 65 65B

[57]

1400000000000 1.4T

[57]

6300[58] Non-commercial research
Facebook's license and distribution scheme restricted access to approved researchers, but the model weights were leaked and became widely available.
Stanford Alpaca 2023-03-13 Stanford University 7 7B

[59]

52000 52K demonstrations


100 <US$100


Non-commercial
Fine-tuned from LLaMA 7B on instruction-following demonstrations generated using OpenAI's text-davinci-003.
GPT-4 2023-03 OpenAI Unknown Unknown Unknown Proprietary
Cerebras-GPT 2023-03 Cerebras 13 13B

[60]

270[28] Apache 2.0
Falcon 2023-03 Technology Innovation Institute 40 40B

[61]

1T tokens, from RefinedWeb (filtered web text corpus)[62] plus some "curated corpora".[63] 2800[58] Apache 2.0[64]
BloombergGPT 2023-03 Bloomberg L.P. 50 50B


363B tokens from Bloomberg's proprietary data sources, plus 345B tokens from general purpose datasets[65] Unknown Unreleased
Designed for financial tasks.[65]
PanGu-Σ 2023-03 Huawei 1085 1085B


329B tokens[66] Proprietary
OpenAssistant[67] 2023-03 LAION 17 17B


1.5T tokens Apache 2.0
Vicuna 2023-03 LMSYS 13 13B


70000 70K conversations

[68]

US$300 Non-commercial
Fine-tuned from LLaMA on user-shared conversations collected from ShareGPT.
Jurassic-2[69][70] 2023-03 AI21 Labs Unknown Unknown Unknown Proprietary
PaLM 2 (Pathways Language Model 2) 2023-05 Google 340 340B

[71]

3600000000000 3.6T 
tokens[71]
85000 85,000

[58]

Used in the Bard chatbot.[72]
YandexGPT 2023-05-17 Yandex Unknown Unknown Unknown
Phi-1 2023-06-21 Microsoft 1.3 1.3B

[73]

7B tokens[73] Unknown MIT
Trained for 4 days on 8 A100s.[73]
Llama 2 2023-07 Meta AI 70 70B

[74]

2000000000000 2T 
tokens[74]
21000 21,000


Llama 2
Trained over 3.3 million GPU (A100) hours.[75]
Claude 2 2023-07 Anthropic Unknown Unknown Unknown Proprietary
Used in the Claude chatbot.[76]
Granite 13b 2023-07 IBM Unknown
Used in IBM Watsonx.[77]
Mistral 7B 2023-09 Mistral AI 7.3 7.3B

[78]

Apache 2.0
YandexGPT 2 2023-09-07 Yandex Unknown Proprietary
Claude 2.1 2023-11 Anthropic Unknown
Used in the Claude chatbot. Has a context window of 200,000 tokens, or ~500 pages.[79]
Grok-1[80] 2023-11 xAI 314 314B


Apache 2.0
Used in the Grok chatbot. Grok 1 has a context length of 8,192 tokens and has access to X (Twitter).[81]
Gemini 1.0 2023-12 Google DeepMind Unknown Proprietary
Multimodal model, comes in three sizes. Used in the chatbot of the same name.[82]
Mixtral 8x7B 2023-12 Mistral AI 46.7 46.7B


Apache 2.0
Outperforms GPT-3.5 and Llama 2 70B on many benchmarks.[83] Mixture of experts model, with 12.9 billion parameters activated per token.[84]
DeepSeek-LLM Template:DTS DeepSeek 67 67B


2T tokens[85]Template:Pg 12000 12,000


DeepSeek
Trained on English and Chinese text. Used 1024 training FLOPs for 67B model, 10b FLOPs for 7B.[85]Template:Pg
Phi-2 2023-12 Microsoft 2.7 2.7B


1.4T tokens 419[86] MIT
Trained on real and synthetic "textbook-quality" data over 14 days on 96 A100 GPUs.[86]

2024

Name Release date[lower-alpha 2] Developer Template:No. of params Corpus size Training cost License[lower-alpha 3] Notes
Gemini 1.5 2024-02 Google DeepMind Unknown Unknown Unknown Proprietary
Multimodal model based on a MoE architecture. Context window above 1 million tokens.[87]
Gemini Ultra 2024-02
Gemma 2024-02 7 7B


6T tokens Gemma Terms of Use[88]
OLMo 2024-02 Allen Institute for AI 7 7B

[89]

2T tokens[90] Apache 2.0[lower-alpha 4]
Claude 3 2024-03 Anthropic Unknown Unknown Proprietary
Includes three models: Haiku, Sonnet, and Opus.[91]
DBRX 2024-03 Databricks and Mosaic ML 136 136B


12T tokens Databricks Open Model[92][93]
YandexGPT 3 Pro 2024-03-28 Yandex Unknown Unknown Proprietary
Fugaku-LLM[94] 2024-05 Fujitsu, Tokyo Institute of Technology, Tohoku University, RIKEN, etc. 13 13B


380B tokens Fugaku-LLM Terms of Use[95]
The largest model ever trained on CPU-only, on the Fugaku supercomputer; the model was trained from scratch on 380 billion tokens using 13,824 Fugaku nodes.[94][96]
Chameleon 2024-05 Meta AI 34 34B

[97]

4400000000000 4.4T


Non-commercial research[98]
Mixtral 8x22B[99] 2024-04-17 Mistral AI 141 141B


Unknown Apache 2.0
Phi-3 2024-04-23 Microsoft 14 14B

[100]

4.8T tokens[101] MIT
Marketed by Microsoft as a "small language model".[100]
Granite Code Models 2024-05 IBM Unknown Unknown Apache 2.0
YandexGPT 3 Lite 2024-05-28 Yandex Unknown Unknown Proprietary
Qwen2 2024-06 Alibaba Cloud 72 72B

[102]

3T tokens Various
DeepSeek-V2 Template:DTS DeepSeek 236 236B


8.1T tokens 28000 28,000


DeepSeek
1.4M hours on H800.[103]
Nemotron-4 2024-06 Nvidia 340 340B


9T tokens 200000 200,000


NVIDIA Open Model[104][105]
Trained for 1 epoch. Trained on 6144 H100 GPUs between December 2023 and May 2024.[106][107]
Claude 3.5 2024-06 Anthropic Unknown Unknown Unknown Proprietary
Initially, only one model, Sonnet, was released.[108] In October 2024, Sonnet 3.5 was upgraded, and Haiku 3.5 became available.[109]
Llama 3.1 2024-07 Meta AI 405 405B


15.6T tokens 440000 440,000


Llama 3
405B version took 31 million hours on H100-80GB, at 3.8E25 FLOPs.[110][111]
Grok-2 2024-08-14 xAI Unknown Unknown Unknown xAI Community License Agreement[112][113]
Originally closed-source, then re-released as "Grok 2.5" under a source-available license in August 2025.[114][115]
OpenAI o1 2024-09-12 OpenAI Unknown Unknown Proprietary
First LLM described as a "reasoning model".[116][117]
Sarvam-1 2024-10-24 Sarvam AI 2 2B


~2T tokens Sarvam AI Research
Supports 10 Indic languages and English[118][119]
YandexGPT 4 Lite and Pro 2024-10-24 Yandex Unknown Unknown Proprietary
Mistral Large 2024-11 Mistral AI 123 123B


Unknown Mistral Research
Upgraded over time. The latest version is 24.11.[120]
Pixtral
Multimodal. There is also a 12B version which is under Apache 2 license.[120]
OLMo 2 2024-11 Allen Institute for AI 32 32B

[121][122]

6.6T tokens[122] 15,000[122] Apache 2.0[lower-alpha 4]
Phi-4 2024-12-12 Microsoft 14 14B

[123]

9800000000000 9.8T 
tokens
Unknown MIT
Marketed by Microsoft as a "small language model".[124]
DeepSeek-V3 2024-12 DeepSeek 671 671B


14.8T tokens 56000 56,000


Used 2.788M training hours on H800 GPUs.[125] Originally released under the DeepSeek License, then re-released under the MIT License as "DeepSeek-V3-0324" in March 2025.[126]
Amazon Nova 2024-12 Amazon Unknown Unknown Unknown Proprietary
Includes three models: Nova Micro, Nova Lite, and Nova Pro.[127]

2025

Name Release date[lower-alpha 2] Developer Template:No. of params Corpus size[lower-alpha 5] License[lower-alpha 3] Notes
DeepSeek-R1 2025-01-20 DeepSeek 671 671B


Unknown MIT
No pretraining; reinforcement-learned upon V3-Base.[128][129]
Qwen2.5 2025-01-26 Alibaba 72 72B


18000 18T


Various
7 dense models with parameter counts from 0.5B to 72B. Alibaba also released 2 MoE variants.[130]
MiniMax-Text-01 2025-01-14 Minimax 456 456B


4700 4.7T


Minimax Model
[131][132]
Gemini 2.0 2025-02-05 Google DeepMind Unknown Unknown Proprietary
Three models released: Flash, Flash-Lite and Pro.[133][134][135]
Grok 3 2025-02-19 xAI
Training cost claimed to be "10x the compute of previous state-of-the-art models".[136]
Claude 3.7 2025-02-24 Anthropic
One model, Sonnet 3.7.[137]
YandexGPT 5 Lite Pretrain and Pro 2025-02-25 Yandex
GPT-4.5 2025-02-27 OpenAI
OpenAI's largest non-reasoning model at the time.[138]
Gemini 2.5 2025-03-25 Google DeepMind
Three models released: Flash, Flash-Lite and Pro.[139]
YandexGPT 5 Lite Instruct 2025-03-31 Yandex
Llama 4 2025-04-05 Meta AI 400 400B


40000 40T


Llama 4
[140][141]
OpenAI o3 and o4-mini 2025-04-16 OpenAI Unknown Unknown Proprietary
Reasoning models.[142]
Qwen3 2025-04-28 Alibaba Cloud 235 235B


36000 36T


Apache 2.0
Multiple sizes, the smallest being 0.6B.[143]
Claude 4 2025-05-22 Anthropic Unknown Unknown Proprietary
Includes two models, Sonnet and Opus.[144]
Sarvam-M 2025-05-23 Sarvam AI 24 24B


Unknown Apache 2.0
Hybrid reasoning model fine-tuned on Mistral Small base; optimized for math, programming, and Indian languages.[145][146]
Grok 4 2025-07-09 xAI Unknown Unknown Proprietary
[147]
Param-1 2025-07-21 BharatGen 2.9 2.9B


5000 5T


Apache 2.0
Corpus is "focus[ed] on India’s linguistic landscape".[148][149]
GLM-4.5 2025-07-29 Z.ai 355 355B


22000 22T


MIT
Released in 355B and 106B sizes.[150] Corpus size was calculated by combining the 15 trillion tokens and the 7 trillion tokens pre-training mix.[151]
GPT-OSS 2025-08-05 OpenAI 117 117B


Unknown Apache 2.0
Released in 20B and 120B sizes.[152]
Claude 4.1 2025-08-05 Anthropic Unknown Unknown Proprietary
Includes one model, Opus.[153]
GPT-5 2025-08-07 OpenAI Unknown Unknown
Includes three models: GPT-5, GPT-5 mini, and GPT-5 nano. GPT-5 is available in ChatGPT and API. It includes reasoning abilities.[154][155]
DeepSeek-V3.1 2025-08-21 DeepSeek 671 671B


15639 15.639T


MIT
Corpus size based on DeepSeek V3 (trained on 14.8T tokens); further trained on 839B tokens from the extension phases (630B + 209B).[156] A hybrid model that can switch between thinking and non-thinking modes.[157]
YandexGPT 5.1 Pro 2025-08-28 Yandex Unknown Unknown Proprietary
Apertus 2025-09-02 ETH Zurich and EPF Lausanne 70 70B


15000 15T


Apache 2.0
The first LLM to be compliant with the Artificial Intelligence Act of the European Union.[158][159]
Claude Sonnet 4.5 2025-09-29 Anthropic Unknown Unknown Proprietary
[160]
GLM-4.6 2025-09-30 Z.ai 357 357B


Apache 2.0
[161][162][163]
Alice AI LLM 1.0 2025-10-28 Yandex Unknown Proprietary
Gemini 3 2025-11-18 Google DeepMind Unknown
Models released: Deep Think and Pro.[164]
Olmo 3[165] 2025-11-20 Allen Institute for AI 32 32B


5900 5.9T


Apache 2.0[lower-alpha 4]
Includes 7B and 32B parameter versions, alongside reasoning and instruction-following models.[166]
Claude Opus 4.5 2025-11-24 Anthropic Unknown Unknown Proprietary
Largest model in the Claude family.[167]
DeepSeek-V3.2 2025-12-01 DeepSeek 685 685B


MIT
Uses a custom DeepSeek Sparse Attention (DSA) mechanism[168][169][170]
GPT 5.2 2025-12-11 OpenAI Unknown Proprietary
It was able to solve an open problem in statistical learning theory that had previously remained unresolved by human researchers.[171]
GLM-4.7 2025-12-22 Z.ai 355 355B


Apache 2.0

2026

Name Release date[lower-alpha 2] Developer Template:No. of params Corpus size[lower-alpha 5] License[lower-alpha 3] Notes
Qwen3-Max-Thinking 2026-01-26 Alibaba Cloud Unknown Unknown Proprietary
Proprietary reasoning model with adaptive tool-use, test-time scaling, and iterative self-reflection.[172]
K2 Think V2 2026-01-27 MBUZAI / IFM 70 70B


12000 12T

[173]

Apache 2.0
[174]
Kimi K2.5 Moonshot AI 1000 1T


15000 15T


Modified MIT[175]
Multimodal MoE with 32B active parameters, derived from Kimi K2.[176] Can use "Agent Swarm" technology to coordinate up to 100 parallel sub-agents.[177][178]
Step 3.5 Flash 2026-02-12 StepFun 196 196B


Unknown Apache 2.0
MoE model with 11B active parameters out of 196B total[179][180][181]
Claude Opus 4.6 2026-02-05 Anthropic Unknown Proprietary
GPT-5.3-Codex 2026-02-05 OpenAI
GLM-5 2026-02-12 Z.ai 744 744B


28500 28.5T


MIT
40B active parameters[182][183]
Qwen3.5 2026-02-15 Alibaba Cloud 397 397B


Unknown Apache 2.0
17B active parameters[184][185][186]
Qwen3.5-Plus Unknown Proprietary
[184][185]
Claude Sonnet 4.6 2026-02-17 Anthropic
Param-2 2026-02-17 BharatGen 17 17B


22000 22T


BharatGen Research[187]
Mixture-of-experts model, successor of Param-1; many more Indic languages are supported. Trained on H100 GPUs for 24 days.[188]
Sarvam-105B 2026-02-18 Sarvam AI 105 105B


12000 12T


Apache 2.0
India's first independently-trained foundation model; has 105B and 30B versions. Based on mixture-of-experts model, using only 10.3B active parameters at a time.[189] Interprets Indic languages and Hinglish.[190][191][192] An early checkpoint of the model was released in January.[193]
Sarvam-30B 30 30B


16000 16T


GPT-5.4 2026-03-05 OpenAI Unknown Unknown Proprietary
Mistral Small 4 2026-03-17 Mistral AI 119 119B


Apache 2.0
MoE model with 6B active parameters out of 119B total[194][195]
MiMo-V2-Pro 2026-03-18 Xiaomi 1000 1T


Proprietary
Mixture-of-experts (MoE) model with more than 1 trillion parameters (43 billion active). Designed for agentic scenarios. Initially available on OpenRouter under the codename "Hunter Alpha" before official release.[196][197]
Qwen3.6-Plus 2026-04-01 Alibaba Cloud Unknown
[198][199]
Gemma 4 2026-04-02 Google DeepMind 31 31B


Apache 2.0
Released in 31B, 26B A4B (3.8B active parameters), E4B (4B effective parameters), and E2B variants[200][201]
GLM-5.1 2026-04-07 Z.ai 744 744B


28500 28.5T


MIT
40B active parameters[183][202][203]
Muse Spark 2026-04-08 Meta Unknown Unknown Proprietary
[204]
Qwen3.6 2026-04-15 Alibaba Cloud 35 35B


Apache 2.0
MoE model with 3B active parameters out of 35B total[205][206]
Claude Opus 4.7 2026-04-16 Anthropic Unknown Proprietary
Kimi K2.6 2026-04-20 Moonshot AI 1000 1T


Modified MIT[207]
32B active parameters[208][209][210]
GPT-5.5 2026-04-23 OpenAI Unknown Proprietary
Hy3 preview 2026-04-23 Tencent 295 295B


Tencent Hy Community[211]
Mixture-of-experts model with 21B active parameters and a 256K-token context window. Preview predecessor to Hy3.[212]
DeepSeek-V4-Flash Template:DTS DeepSeek 284 284B


32000 32T


MIT
13B (Flash) and 49B (Pro) active parameters.[213][214] Preview release on 24 April.[215] Official releases of V4-Flash on 31 July,[216], and V4-Pro on 13 August.[213]
DeepSeek-V4-Pro 1600 1.6T


MiMo-V2.5-Pro 2026-04-27 Xiaomi 1020 1.02T


48000 48T


MoE model designed for agentic coding and long-horizon software engineering tasks.[217][218]
MiMo-V2.5 310 310B


27000 27T


Omni-modal MoE model with agentic capabilities and 1M-token context.[219]
Mistral Medium 3.5 2026-04-28 Mistral AI 128B Unknown Modified MIT[220]
Mistral AI's flagship model that merges instruction-following, reasoning, and coding into a single 128B dense model[221]
Composer 2.5 2026-05-18 Cursor 1000 1T


Proprietary
Agentic coding model available in the Cursor IDE. Built on Kimi K2.5 with additional reinforcement learning; reported 1T total / 32B active parameters.[222]
Gemini 3.5 Flash 2026-05-19 Google DeepMind Unknown
[223]
Qwen3.7-Max 2026-05-20 Alibaba Cloud
[224][225]
Claude Opus 4.8 2026-05-28 Anthropic
[226]
Step 3.7 Flash 2026-05-29 StepFun 198 198B


Apache 2.0
Parameter count is 196B (base) + 1.8B (ViT).[227]
MiniMax-M3 2026-06-01 MiniMax 428 428B


MiniMax Community[228]
23B active parameters[229][230]
Nemotron 3 Ultra 2026-06-04 Nvidia 550 550B


20000 20T


OpenMDW 1.1[231]
55B active parameters[232][233]
Claude Fable 5 2026-06-09 Anthropic Unknown Unknown Proprietary
[234]
Kimi K2.7 Code 2026-06-16 Moonshot AI 1000 1T


Modified MIT[235]
32B active parameters[236][237]
GLM-5.2 2026-06-16 Z.ai 744 744B


28500 28.5T


MIT
40B active parameters[238][239][240]
LongCat-2.0 2026-06-30 Meituan 1600 1.6T


35000 35T


48B active parameters. Preview version previously released as "Owl Alpha" on OpenRouter.[241][242]
Hy3 2026-07-06 Tencent 295 295B


Unknown Apache 2.0
21B active parameters[243]
Grok 4.5 2026-07-08 SpaceXAI Unknown Proprietary
Frontier model for coding, agentic tasks, and knowledge work; trained alongside Cursor. Public availability followed on 9 July 2026.[244][245]
SWE-1.7 2026-07-08 Cognition AI
Coding model for the Devin agent; trained from a Kimi K2.7 base with Cognition's reinforcement-learning pipeline. Available via Devin (not as open weights).[246]
Muse Spark 1.1 2026-07-09 Meta
[247]
GPT-5.6 2026-07-09 OpenAI
Inkling 2026-07-15 Thinking Machines Lab 975 975B


45000 45T


Apache 2.0
41B active parameters[248][249]
Kimi K3 2026-07-16 Moonshot AI 2800 2.8T


Unknown Kimi K3[250]
104B active parameters. Weights released on 27 July.[251][252]
Qwen3.8 2026-07-19 Alibaba Cloud 2400 2.4T


Various
Available in 2.4T (95B active parameters) and 27B variants.[253][254] Preview of Qwen3.8-Max (proprietary variant with image input) was released on 19 July.[255] Weights for 2.4T variant released on 12 August.[253] 27B variant released on 14 August.[254]
Gemini 3.5 Flash Lite Template:DTS Google DeepMind Unknown Proprietary
[256]
Gemini 3.6 Flash Template:DTS Google DeepMind
[256]
Laguna S 2.1 2026-07-21 Poolside AI 118 118B


OpenMDW 1.1[231]
8B active parameters[257][258]
Motif 3 2026-07-21 Motif Technologies 314 314B


12500 12.5T


MIT
13.2B active parameters. Preview release on 21 July.[259] Official version and weights released on 13 August.[260][261]
Solar Open 2 2026-07-23 Upstage 250 250B


12000 12T


Upstage Solar[262]
15B active parameters[263][264]
Fugu-Ultra v1.1 2026-07-24 Sakana AI Unknown Unknown Proprietary
Upgraded Fugu-Ultra with stronger coding, agentic, and reasoning benchmarks vs v1.0.[265]
Claude Opus 5 2026-07-24 Anthropic
[266]
A.X K2 2026-07-29 SK Telecom 688 688B


8200 8.2T


Apache 2.0
33B active parameters[267][268]
Inkling Small 2026-07-31 Thinking Machines Lab 276 276B


Unknown
12B active parameters[269]
K-Exaone 2.0 2026-07-31 LG 750 750B


37B active parameters[270][271]
Muse Glimmer 2026-08-10 Meta 30 30B


[272]
Gemini 3.7 Flash Template:DTS Google DeepMind Unknown Proprietary
[273]
GLM-5.3 2026-08-14 Z.ai 744 744B


30000 30T


Various 40B active
GLM-5.3-flash 28 August 320B 30T MIT 18B active
Hy4 preview 2026-08-28 Tencent 770 770B


Unknown Apache 2.0
[274] 49B active parameters. Preview release.[275][276]
Claude Mythos 5.1 2026-09-01 Anthropic Unknown Proprietary
[277]
Claude Fable 5.1 2026-09-01 Anthropic
[277]
Gemini 3.8 Flash Template:DTS Google DeepMind
[278]
GPT-6 Astra 2026-09-03 OpenAI
[279]
K2 Horizon MBUZAI / IFM 0.9B to 375B 20000 20T


Apache 2.0[lower-alpha 4]
A family of six different models[280]
DeepSeek-V4.1-Flash 2026-09-10 DeepSeek 552 552B


45000 45T


MIT
8B (prefill) and 16B (decode) active parameters.[281][282]
Grok 4.7 2026-09-21 SpaceXAI Unknown Unknown Proprietary
[283]
GPT-6 Sol 2026-09-22 OpenAI
[284]
MiMo-V2.6-Pro Xiaomi 1020 1.02T

[285]

30000 30T

[285]

MIT
Native omni-modal models released as part of the MiMo-V2.6 series.[286]
MiMo-V2.6-Flash 310 310B

[285]

48000 48T

[285]

Claude Opus 5.5 Anthropic Unknown Unknown Proprietary
[287]
Claude Sonnet 5.5 2026-09-28
[288]
GPT 6.1 2026-09-29 OpenAI
[289]
Gemini 4 Argon 2026-09-30 Google DeepMind
[290]

See also

Notes

  1. ↑ In many cases, researchers release or report on multiple versions of a model having different sizes. In these cases, the size of the largest model is listed here.
  2. ↑ 2.0 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 This is the date that documentation describing the model's architecture was first released.
  3. ↑ 3.0 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 License applies only to the pre-trained model weights unless otherwise noted. In many cases, the training code, datasets and post-training data remain proprietary. LLMs may be licensed differently from the chatbots that use them; for the licenses of chatbots, see List of chatbots.
  4. ↑ 4.0 4.1 4.2 4.3 Stack is fully open-source, including training code, datasets and post-training data.
  5. ↑ 5.0 5.1 Expressed in number of tokens, unless otherwise specified

References

  1. ↑ "Improving language understanding with unsupervised learning". 11 June 2018. https://openai.com/research/language-unsupervised. 
  2. ↑ "finetune-transformer-lm". GitHub. https://github.com/openai/finetune-transformer-lm. 
  3. ↑ Radford, Alec (11 June 2018). "Improving language understanding with unsupervised learning". https://openai.com/index/language-unsupervised/. 
  4. ↑ 4.0 4.1 Devlin, Jacob; Chang, Ming-Wei; Lee, Kenton; Toutanova, Kristina (11 October 2018). "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding". arXiv:1810.04805v2 [cs.CL].
  5. ↑ Prickett, Nicole Hemsoth (24 August 2021). "Cerebras Shifts Architecture To Meet Massive AI/ML Models". https://www.nextplatform.com/2021/08/24/cerebras-shifts-architecture-to-meet-massive-ai-ml-models/. 
  6. ↑ "BERT". 13 March 2023. https://github.com/google-research/bert. 
  7. ↑ 7.0 7.1 Raffel, Colin; Shazeer, Noam; Roberts, Adam; Lee, Katherine; Narang, Sharan; Matena, Michael; Zhou, Yanqi; Li, Wei et al. (2020). "Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer". Journal of Machine Learning Research 21 (140): 1–67. ISSN 1533-7928. http://jmlr.org/papers/v21/20-074.html. 
  8. ↑ google-research/text-to-text-transfer-transformer, Google Research, 2 April 2024, https://github.com/google-research/text-to-text-transfer-transformer, retrieved 4 April 2024 
  9. ↑ "Imagen: Text-to-Image Diffusion Models". https://imagen.research.google/. 
  10. ↑ "Pretrained models — transformers 2.0.0 documentation". https://huggingface.co/transformers/v2.0.0/pretrained_models.html. 
  11. ↑ "xlnet". GitHub. https://github.com/zihangdai/xlnet/. 
  12. ↑ Yang, Zhilin; Dai, Zihang; Yang, Yiming; Carbonell, Jaime; Salakhutdinov, Ruslan; Le, Quoc V. (2 January 2020). "XLNet: Generalized Autoregressive Pretraining for Language Understanding". arXiv:1906.08237 [cs.CL].
  13. ↑ "GPT-2: 1.5B Release". 5 November 2019. https://openai.com/blog/gpt-2-1-5b-release/. 
  14. ↑ "Better language models and their implications". https://openai.com/research/better-language-models. 
  15. ↑ 15.0 15.1 "OpenAI's GPT-3 Language Model: A Technical Overview". 3 June 2020. https://lambdalabs.com/blog/demystifying-gpt-3. 
  16. ↑ 16.0 16.1 "openai-community/gpt2-xl · Hugging Face". 21 October 2019. https://huggingface.co/openai-community/gpt2-xl. 
  17. ↑ "gpt-2". GitHub. https://github.com/openai/gpt-2. 
  18. ↑ Wiggers, Kyle (28 April 2022). "The emerging types of language models and why they matter". TechCrunch. https://techcrunch.com/2022/04/28/the-emerging-types-of-language-models-and-why-they-matter/. 
  19. ↑ Table D.1 in Brown, Tom B.; Mann, Benjamin; Ryder, Nick; Subbiah, Melanie; Kaplan, Jared; Dhariwal, Prafulla; Neelakantan, Arvind; Shyam, Pranav; Sastry, Girish; Askell, Amanda; Agarwal, Sandhini; Herbert-Voss, Ariel; Krueger, Gretchen; Henighan, Tom; Child, Rewon; Ramesh, Aditya; Ziegler, Daniel M.; Wu, Jeffrey; Winter, Clemens; Hesse, Christopher; Chen, Mark; Sigler, Eric; Litwin, Mateusz; Gray, Scott; Chess, Benjamin; Clark, Jack; Berner, Christopher; McCandlish, Sam; Radford, Alec; Sutskever, Ilya; Amodei, Dario (28 May 2020). "Language Models are Few-Shot Learners". arXiv:2005.14165v4 [cs.CL].
  20. ↑ "ChatGPT: Optimizing Language Models for Dialogue". 30 November 2022. https://openai.com/blog/chatgpt/. 
  21. ↑ "GPT Neo". 15 March 2023. https://github.com/EleutherAI/gpt-neo. 
  22. ↑ 22.0 22.1 22.2 Gao, Leo; Biderman, Stella; Black, Sid; Golding, Laurence; Hoppe, Travis; Foster, Charles; Phang, Jason; He, Horace; Thite, Anish; Nabeshima, Noa; Presser, Shawn; Leahy, Connor (31 December 2020). "The Pile: An 800GB Dataset of Diverse Text for Language Modeling". arXiv:2101.00027 [cs.CL].
  23. ↑ 23.0 23.1 Iyer, Abhishek (15 May 2021). "GPT-3's free alternative GPT-Neo is something to be excited about". VentureBeat. https://venturebeat.com/ai/gpt-3s-free-alternative-gpt-neo-is-something-to-be-excited-about/. 
  24. ↑ Kim, Han-joo (25 May 2021). "Naver unveils supersized AI platform HyperCLOVA". https://en.yna.co.kr/view/AEN20210525005400320. 
  25. ↑ 25.0 25.1 Wiggers, Kyle (1 June 2021). "Naver trained a 'GPT-3-like' Korean language model". https://venturebeat.com/ai/naver-trained-a-gpt-3-like-korean-language-model. 
  26. ↑ 26.0 26.1 Kim, Boseop; Kim, HyoungSeok; Lee, Sang-Woo; Lee, Gichang; Kwak, Donghyun; Dong Hyeon, Jeon; Park, Sunghyun; Kim, Sungju et al. (November 2021). "What Changes Can Large-scale Language Models Bring? Intensive Study on HyperCLOVA: Billions-scale Korean Generative Pretrained Transformers". in Moens, Marie-Francine; Huang, Xuanjing; Specia, Lucia et al.. Online and Punta Cana, Dominican Republic: Association for Computational Linguistics. pp. 3405–3424. doi:10.18653/v1/2021.emnlp-main.274. https://aclanthology.org/2021.emnlp-main.274/. 
  27. ↑ "GPT-J-6B: An Introduction to the Largest Open Source GPT Model | Forefront". https://www.forefront.ai/blog-posts/gpt-j-6b-an-introduction-to-the-largest-open-sourced-gpt-model. 
  28. ↑ 28.0 28.1 28.2 28.3 Dey, Nolan; Gosal, Gurpreet; Zhiming; Chen; Khachane, Hemant; Marshall, William; Pathria, Ribhu; Tom, Marvin; Hestness, Joel (1 April 2023). "Cerebras-GPT: Open Compute-Optimal Language Models Trained on the Cerebras Wafer-Scale Cluster". arXiv:2304.03208 [cs.LG].
  29. ↑ Alvi, Ali; Kharya, Paresh (11 October 2021). "Using DeepSpeed and Megatron to Train Megatron-Turing NLG 530B, the World's Largest and Most Powerful Generative Language Model". https://www.microsoft.com/en-us/research/blog/using-deepspeed-and-megatron-to-train-megatron-turing-nlg-530b-the-worlds-largest-and-most-powerful-generative-language-model/. 
  30. ↑ 30.0 30.1 Smith, Shaden; Patwary, Mostofa; Norick, Brandon; LeGresley, Patrick; Rajbhandari, Samyam; Casper, Jared; Liu, Zhun; Prabhumoye, Shrimai; Zerveas, George; Korthikanti, Vijay; Zhang, Elton; Child, Rewon; Aminabadi, Reza Yazdani; Bernauer, Julie; Song, Xia (4 February 2022). "Using DeepSpeed and Megatron to Train Megatron-Turing NLG 530B, A Large-Scale Generative Language Model". arXiv:2201.11990 [cs.CL].
  31. ↑ 31.0 31.1 Rajbhandari, Samyam; Li, Conglong; Yao, Zhewei; Zhang, Minjia; Aminabadi, Reza Yazdani; Awan, Ammar Ahmad; Rasley, Jeff; He, Yuxiong (21 July 2022), DeepSpeed-MoE: Advancing Mixture-of-Experts Inference and Training to Power Next-Generation AI Scale 
  32. ↑ Wang, Shuohuan; Sun, Yu; Xiang, Yang; Wu, Zhihua; Ding, Siyu; Gong, Weibao; Feng, Shikun; Shang, Junyuan; Zhao, Yanbin; Pang, Chao; Liu, Jiaxiang; Chen, Xuyi; Lu, Yuxiang; Liu, Weixin; Wang, Xi; Bai, Yangfan; Chen, Qiuliang; Zhao, Li; Li, Shiyong; Sun, Peng; Yu, Dianhai; Ma, Yanjun; Tian, Hao; Wu, Hua; Wu, Tian; Zeng, Wei; Li, Ge; Gao, Wen; Wang, Haifeng (23 December 2021). "ERNIE 3.0 Titan: Exploring Larger-scale Knowledge Enhanced Pre-training for Language Understanding and Generation". arXiv:2112.12731 [cs.CL].
  33. ↑ "Product". https://www.anthropic.com/product. 
  34. ↑ 34.0 34.1 Askell, Amanda; Bai, Yuntao; Chen, Anna; et al. (9 December 2021). "A General Language Assistant as a Laboratory for Alignment". arXiv:2112.00861 [cs.CL].
  35. ↑ Bai, Yuntao; Kadavath, Saurav; Kundu, Sandipan; et al. (15 December 2022). "Constitutional AI: Harmlessness from AI Feedback". arXiv:2212.08073 [cs.CL].
  36. ↑ 36.0 36.1 36.2 Dai, Andrew M; Du, Nan (9 December 2021). "More Efficient In-Context Learning with GLaM". https://ai.googleblog.com/2021/12/more-efficient-in-context-learning-with.html. 
  37. ↑ "Language modelling at scale: Gopher, ethical considerations, and retrieval". 8 December 2021. https://www.deepmind.com/blog/language-modelling-at-scale-gopher-ethical-considerations-and-retrieval. 
  38. ↑ 38.0 38.1 38.2 Hoffmann, Jordan; Borgeaud, Sebastian; Mensch, Arthur; et al. (29 March 2022). "Training Compute-Optimal Large Language Models". arXiv:2203.15556 [cs.CL].
  39. ↑ 39.0 39.1 39.2 39.3 Table 20 and page 66 of PaLM: Scaling Language Modeling with Pathways
  40. ↑ 40.0 40.1 Cheng, Heng-Tze; Thoppilan, Romal (21 January 2022). "LaMDA: Towards Safe, Grounded, and High-Quality Dialog Models for Everything". https://ai.googleblog.com/2022/01/lamda-towards-safe-grounded-and-high.html. 
  41. ↑ Thoppilan, Romal; De Freitas, Daniel; Hall, Jamie; Shazeer, Noam; Kulshreshtha, Apoorv; Cheng, Heng-Tze; Jin, Alicia; Bos, Taylor; Baker, Leslie; Du, Yu; Li, YaGuang; Lee, Hongrae; Zheng, Huaixiu Steven; Ghafouri, Amin; Menegali, Marcelo (1 January 2022). "LaMDA: Language Models for Dialog Applications". arXiv:2201.08239 [cs.CL].
  42. ↑ Black, Sidney; Biderman, Stella; Hallahan, Eric (1 May 2022). "GPT-NeoX-20B: An Open-Source Autoregressive Language Model". Proceedings of BigScience Episode #5 – Workshop on Challenges & Perspectives in Creating Large Language Models. Proceedings of BigScience Episode #5 – Workshop on Challenges & Perspectives in Creating Large Language Models. pp. 95–136. https://aclanthology.org/2022.bigscience-1.9/. Retrieved 19 December 2022. 
  43. ↑ 43.0 43.1 43.2 Hoffmann, Jordan; Borgeaud, Sebastian; Mensch, Arthur; Sifre, Laurent (12 April 2022). "An empirical analysis of compute-optimal large language model training". Deepmind Blog. https://www.deepmind.com/blog/an-empirical-analysis-of-compute-optimal-large-language-model-training. 
  44. ↑ Narang, Sharan; Chowdhery, Aakanksha (4 April 2022). "Pathways Language Model (PaLM): Scaling to 540 Billion Parameters for Breakthrough Performance". https://ai.googleblog.com/2022/04/pathways-language-model-palm-scaling-to.html. 
  45. ↑ Zhang, Susan; Diab, Mona; Zettlemoyer, Luke. "Democratizing access to large-scale language models with OPT-175B". https://ai.facebook.com/blog/democratizing-access-to-large-scale-language-models-with-opt-175b/. 
  46. ↑ Zhang, Susan; Roller, Stephen; Goyal, Naman; Artetxe, Mikel; Chen, Moya; Chen, Shuohui; Dewan, Christopher; Diab, Mona; Li, Xian; Lin, Xi Victoria; Mihaylov, Todor; Ott, Myle; Shleifer, Sam; Shuster, Kurt; Simig, Daniel; Koura, Punit Singh; Sridhar, Anjali; Wang, Tianlu; Zettlemoyer, Luke (21 June 2022). "OPT: Open Pre-trained Transformer Language Models". arXiv:2205.01068 [cs.CL].
  47. ↑ "metaseq/projects/OPT/chronicles at main · facebookresearch/metaseq". https://github.com/facebookresearch/metaseq/tree/main/projects/OPT/chronicles. 
  48. ↑ 48.0 48.1 Khrushchev, Mikhail; Vasilev, Ruslan; Petrov, Alexey; Zinov, Nikolay (22 June 2022), YaLM 100B, https://github.com/yandex/YaLM-100B, retrieved 18 March 2023 
  49. ↑ 49.0 49.1 Lewkowycz, Aitor; Andreassen, Anders; Dohan, David; Dyer, Ethan; Michalewski, Henryk; Ramasesh, Vinay; Slone, Ambrose; Anil, Cem; Schlag, Imanol; Gutman-Solo, Theo; Wu, Yuhuai; Neyshabur, Behnam; Gur-Ari, Guy; Misra, Vedant (30 June 2022). "Solving Quantitative Reasoning Problems with Language Models". arXiv:2206.14858 [cs.CL].
  50. ↑ "Minerva: Solving Quantitative Reasoning Problems with Language Models". 30 June 2022. https://ai.googleblog.com/2022/06/minerva-solving-quantitative-reasoning.html. 
  51. ↑ Ananthaswamy, Anil (8 March 2023). "In AI, is bigger always better?". Nature 615 (7951): 202–205. doi:10.1038/d41586-023-00641-w. PMID 36890378. Bibcode: 2023Natur.615..202A. https://www.nature.com/articles/d41586-023-00641-w. Retrieved 9 March 2023. 
  52. ↑ "bigscience/bloom · Hugging Face". 9 November 2022. https://huggingface.co/bigscience/bloom. 
  53. ↑ Taylor, Ross; Kardas, Marcin; Cucurull, Guillem; Scialom, Thomas; Hartshorn, Anthony; Saravia, Elvis; Poulton, Andrew; Kerkez, Viktor; Stojnic, Robert (16 November 2022). "Galactica: A Large Language Model for Science". arXiv:2211.09085 [cs.CL].
  54. ↑ "20B-parameter Alexa model sets new marks in few-shot learning". 2 August 2022. https://www.amazon.science/blog/20b-parameter-alexa-model-sets-new-marks-in-few-shot-learning. 
  55. ↑ Soltan, Saleh; Ananthakrishnan, Shankar; FitzGerald, Jack; et al. (3 August 2022). "AlexaTM 20B: Few-Shot Learning Using a Large-Scale Multilingual Seq2Seq Model". arXiv:2208.01448 [cs.CL].
  56. ↑ "AlexaTM 20B is now available in Amazon SageMaker JumpStart | AWS Machine Learning Blog". 17 November 2022. https://aws.amazon.com/blogs/machine-learning/alexatm-20b-is-now-available-in-amazon-sagemaker-jumpstart/. 
  57. ↑ 57.0 57.1 "Introducing LLaMA: A foundational, 65-billion-parameter large language model". Meta AI. 24 February 2023. https://ai.facebook.com/blog/large-language-model-llama-meta-ai/. 
  58. ↑ 58.0 58.1 58.2 "The Falcon has landed in the Hugging Face ecosystem". 19 May 2023. https://huggingface.co/blog/falcon. 
  59. ↑ Taori, Rohan; Gulrajani, Ishaan; Zhang, Tianyi; Dubois, Yann; Li, Xuechen; Guestrin, Carlos; Liang, Percy; Hashimoto, Tatsunori B. (13 March 2023). "Alpaca: A Strong, Replicable Instruction-Following Model". Stanford University. https://crfm.stanford.edu/2023/03/13/alpaca.html. 
  60. ↑ Dey, Nolan (28 March 2023). "Cerebras-GPT: A Family of Open, Compute-efficient, Large Language Models". https://www.cerebras.net/blog/cerebras-gpt-a-family-of-open-compute-efficient-large-language-models/. 
  61. ↑ "Abu Dhabi-based TII launches its own version of ChatGPT". https://fastcompanyme.com/news/abu-dhabi-based-tii-launches-its-own-version-of-chatgpt/. 
  62. ↑ Penedo, Guilherme; Malartic, Quentin; Hesslow, Daniel; Cojocaru, Ruxandra; Cappelli, Alessandro; Alobeidli, Hamza; Pannier, Baptiste; Almazrouei, Ebtesam; Launay, Julien (1 June 2023). "The RefinedWeb Dataset for Falcon LLM: Outperforming Curated Corpora with Web Data, and Web Data Only". arXiv:2306.01116 [cs.CL].
  63. ↑ "tiiuae/falcon-40b · Hugging Face". 9 June 2023. https://huggingface.co/tiiuae/falcon-40b. 
  64. ↑ UAE's Falcon 40B, World's Top-Ranked AI Model from Technology Innovation Institute, is Now Royalty-Free , 31 May 2023
  65. ↑ 65.0 65.1 Wu, Shijie; Irsoy, Ozan; Lu, Steven; Dabravolski, Vadim; Dredze, Mark; Gehrmann, Sebastian; Kambadur, Prabhanjan; Rosenberg, David; Mann, Gideon (30 March 2023). "BloombergGPT: A Large Language Model for Finance". arXiv:2303.17564 [cs.LG].
  66. ↑ Ren, Xiaozhe; Zhou, Pingyi; Meng, Xinfan; Huang, Xinjing; Wang, Yadao; Wang, Weichao; Li, Pengfei; Zhang, Xiaoda; Podolskiy, Alexander; Arshinov, Grigory; Bout, Andrey; Piontkovskaya, Irina; Wei, Jiansheng; Jiang, Xin; Su, Teng; Liu, Qun; Yao, Jun (19 March 2023). "PanGu-Σ: Towards Trillion Parameter Language Model with Sparse Heterogeneous Computing". arXiv:2303.10845 [cs.CL].
  67. ↑ Köpf, Andreas; Kilcher, Yannic; von Rütte, Dimitri; Anagnostidis, Sotiris; Tam, Zhi-Rui; Stevens, Keith; Barhoum, Abdullah; Duc, Nguyen Minh; Stanley, Oliver; Nagyfi, Richárd; ES, Shahul; Suri, Sameer; Glushkov, David; Dantuluri, Arnav; Maguire, Andrew (14 April 2023). "OpenAssistant Conversations – Democratizing Large Language Model Alignment". arXiv:2304.07327 [cs.CL].
  68. ↑ Chiang, Wei-Lin; Li, Zhuohan; Lin, Zi; Sheng, Ying; Wu, Zhanghao; Zhang, Hao; Zheng, Lianmin; Zhuang, Siyuan et al. (30 March 2023). "Vicuna: An Open-Source Chatbot Impressing GPT-4 with 90%* ChatGPT Quality". https://lmsys.org/blog/2023-03-30-vicuna/. 
  69. ↑ Wrobel, Sharon. "Tel Aviv startup rolls out new advanced AI language model to rival OpenAI". ISSN 0040-7909. https://www.timesofisrael.com/ai21-labs-rolls-out-new-advanced-ai-language-model-to-rival-openai/. 
  70. ↑ Wiggers, Kyle (13 April 2023). "With Bedrock, Amazon enters the generative AI race". https://techcrunch.com/2023/04/13/with-bedrock-amazon-enters-the-generative-ai-race/. 
  71. ↑ 71.0 71.1 Elias, Jennifer (16 May 2023). "Google's newest A.I. model uses nearly five times more text data for training than its predecessor". CNBC. https://www.cnbc.com/2023/05/16/googles-palm-2-uses-nearly-five-times-more-text-data-than-predecessor.html. 
  72. ↑ "Introducing PaLM 2". 10 May 2023. https://blog.google/technology/ai/google-palm-2-ai-large-language-model/. 
  73. ↑ 73.0 73.1 73.2 Gunasekar, Suriya; Zhang, Yi; Aneja, Jyoti; Caio César Teodoro Mendes; Allie Del Giorno; Gopi, Sivakanth; Javaheripi, Mojan; Kauffmann, Piero; Gustavo de Rosa; Saarikivi, Olli; Salim, Adil; Shah, Shital; Harkirat Singh Behl; Wang, Xin; Bubeck, Sébastien; Eldan, Ronen; Adam Tauman Kalai; Yin Tat Lee; Li, Yuanzhi (2023). "Textbooks Are All You Need". arXiv:2306.11644 [cs.CL].
  74. ↑ 74.0 74.1 "Introducing Llama 2: The Next Generation of Our Open Source Large Language Model". 2023. https://ai.meta.com/llama/. 
  75. ↑ "llama/MODEL_CARD.md at main · meta-llama/llama". https://github.com/meta-llama/llama/blob/main/MODEL_CARD.md. 
  76. ↑ "Claude 2". https://www.anthropic.com/index/claude-2. 
  77. ↑ Nirmal, Dinesh (7 September 2023). "Building AI for business: IBM's Granite foundation models" (in en-US). https://www.ibm.com/blog/building-ai-for-business-ibms-granite-foundation-models. 
  78. ↑ "Announcing Mistral 7B". 2023. https://mistral.ai/news/announcing-mistral-7b/. 
  79. ↑ "Introducing Claude 2.1". https://www.anthropic.com/index/claude-2-1. 
  80. ↑ xai-org/grok-1, xai-org, 19 March 2024, https://github.com/xai-org/grok-1, retrieved 19 March 2024 
  81. ↑ "Grok-1 model card". https://x.ai/model-card/. 
  82. ↑ "Gemini – Google DeepMind". https://deepmind.google/technologies/gemini/#capabilities. 
  83. ↑ Franzen, Carl (11 December 2023). "Mistral shocks AI community as latest open source model eclipses GPT-3.5 performance". https://venturebeat.com/ai/mistral-shocks-ai-community-as-latest-open-source-model-eclipses-gpt-3-5-performance/. 
  84. ↑ "Mixtral of experts". 11 December 2023. https://mistral.ai/news/mixtral-of-experts/. 
  85. ↑ 85.0 85.1 DeepSeek-AI; Bi, Xiao; Chen, Deli; Chen, Guanting; Chen, Shanhuang; Dai, Damai; Deng, Chengqi; Ding, Honghui; Dong, Kai (5 January 2024). "DeepSeek LLM: Scaling Open-Source Language Models with Longtermism". arXiv:2401.02954 [cs.CL].
  86. ↑ 86.0 86.1 Hughes, Alyssa (12 December 2023). "Phi-2: The surprising power of small language models". https://www.microsoft.com/en-us/research/blog/phi-2-the-surprising-power-of-small-language-models/. 
  87. ↑ "Our next-generation model: Gemini 1.5". 15 February 2024. https://blog.google/technology/ai/google-gemini-next-generation-model-february-2024/#context-window. "This means 1.5 Pro can process vast amounts of information in one go — including 1 hour of video, 11 hours of audio, codebases with over 30,000 lines of code or over 700,000 words. In our research, we’ve also successfully tested up to 10 million tokens." 
  88. ↑ "Gemma". https://ai.google.dev/gemma/terms. 
  89. ↑ "OLMo: Open Language Model | Ai2". https://allenai.org/blog/olmo-open-language-model-87ccfc95f580. 
  90. ↑ Groeneveld, Dirk; Beltagy, Iz; Walsh, Pete; Bhagia, Akshita; Kinney, Rodney; Tafjord, Oyvind; Jha, Ananya Harsh; Ivison, Hamish; Magnusson, Ian (7 June 2024). "OLMo: Accelerating the Science of Language Models". arXiv:2402.00838 [cs.CL].
  91. ↑ "Introducing the next generation of Claude". https://www.anthropic.com/news/claude-3-family. 
  92. ↑ "Databricks Open Model License". 27 March 2024. https://www.databricks.com/legal/open-model-license. 
  93. ↑ "Databricks Open Model Acceptable Use Policy". 27 March 2024. https://www.databricks.com/legal/acceptable-use-policy-open-model. 
  94. ↑ 94.0 94.1 "Release of "Fugaku-LLM" - a large language model trained on the supercomputer "Fugaku"". 10 May 2024. https://info.archives.global.fujitsu/global/about/resources/news/press-releases/2024/0510-01.html. 
  95. ↑ "Fugaku-LLM Terms of Use". 23 April 2024. https://huggingface.co/Fugaku-LLM/Fugaku-LLM-13B/blob/main/LICENSE. 
  96. ↑ "Fugaku-LLM/Fugaku-LLM-13B · Hugging Face". https://huggingface.co/Fugaku-LLM/Fugaku-LLM-13B. 
  97. ↑ Dickson, Ben (22 May 2024). "Meta introduces Chameleon, a state-of-the-art multimodal model". VentureBeat. https://venturebeat.com/ai/meta-introduces-chameleon-a-state-of-the-art-multimodal-model/. 
  98. ↑ "chameleon/LICENSE at e3b711ef63b0bb3a129cf0cf0918e36a32f26e2c · facebookresearch/chameleon". Meta Research. https://github.com/facebookresearch/chameleon/blob/e3b711ef63b0bb3a129cf0cf0918e36a32f26e2c/LICENSE. 
  99. ↑ AI, Mistral (17 April 2024). "Cheaper, Better, Faster, Stronger". https://mistral.ai/news/mixtral-8x22b/. 
  100. ↑ 100.0 100.1 Bilenko, Misha (23 April 2024). "Introducing Phi-3: Redefining what's possible with SLMs". https://azure.microsoft.com/en-us/blog/introducing-phi-3-redefining-whats-possible-with-slms/. 
  101. ↑ Abdin, Marah; et al. (2024). "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone". arXiv:2404.14219 [cs.CL].
  102. ↑ "Qwen2". https://github.com/QwenLM/Qwen2?spm=a3c0i.28768018.7084722650.1.5cd35c10NEqBXm&file=Qwen1.5. 
  103. ↑ DeepSeek-AI; Liu, Aixin; Feng, Bei; Wang, Bin; Wang, Bingxuan; Liu, Bo; Zhao, Chenggang; Dengr, Chengqi; Ruan, Chong (19 June 2024). "DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model". arXiv:2405.04434 [cs.CL].
  104. ↑ "NVIDIA Open Models License". 16 June 2025. https://www.nvidia.com/en-us/agreements/enterprise-software/nvidia-open-model-license/. 
  105. ↑ "Trustworthy AI". 27 June 2024. https://www.nvidia.com/en-us/agreements/trustworthy-ai/terms/. 
  106. ↑ "nvidia/Nemotron-4-340B-Base · Hugging Face". 14 June 2024. https://huggingface.co/nvidia/Nemotron-4-340B-Base. 
  107. ↑ "Nemotron-4 340B | Research". 14 June 2024. https://research.nvidia.com/publication/2024-06_nemotron-4-340b. 
  108. ↑ "Introducing Claude 3.5 Sonnet". https://www.anthropic.com/news/claude-3-5-sonnet. 
  109. ↑ "Introducing computer use, a new Claude 3.5 Sonnet, and Claude 3.5 Haiku". https://www.anthropic.com/news/3-5-models-and-computer-use. 
  110. ↑ "The Llama 3 Herd of Models" (July 23, 2024) Llama Team, AI @ Meta
  111. ↑ "llama-models/models/llama3_1/MODEL_CARD.md at main · meta-llama/llama-models". https://github.com/meta-llama/llama-models/blob/main/models/llama3_1/MODEL_CARD.md. 
  112. ↑ "LICENSE · xai-org/grok-2 at main". 5 November 2025. https://huggingface.co/xai-org/grok-2/blob/main/LICENSE. 
  113. ↑ "xAI Acceptable Use Policy". 2 January 2025. https://x.ai/legal/acceptable-use-policy. 
  114. ↑ Weatherbed, Jess (14 August 2024). "xAI's new Grok-2 chatbots bring AI image generation to X". https://www.theverge.com/2024/8/14/24220127/grok-ai-chatbot-beta-image-generation-x-xai-update. 
  115. ↑ Ha, Anthony (24 August 2025). "Elon Musk says xAI has open sourced Grok 2.5". https://techcrunch.com/2025/08/24/elon-musk-says-xai-has-open-sourced-grok-2-5/. 
  116. ↑ "Introducing OpenAI o1". https://openai.com/o1/. 
  117. ↑ Paul, Katie; Tong, Anna (13 September 2024). "OpenAI launches new series of AI models with 'reasoning' abilities". https://www.reuters.com/technology/artificial-intelligence/openai-launches-new-series-ai-models-solve-hard-problems-2024-09-12/. 
  118. ↑ Jindal, Siddharth (24 October 2024). "Sarvam AI Launches Sarvam-1, Outperforms Gemma-2 and Llama-3.2". https://analyticsindiamag.com/ai-news-updates/sarvam-ai-launches-sarvam-1-outperforms-gemma-2-and-llama-3-2/. 
  119. ↑ "LICENSE.md · sarvamai/sarvam-1". 23 October 2024. https://huggingface.co/sarvamai/sarvam-1/blob/d3880226af5d8adffd44250463f31ae6fe16073b/LICENSE.md. 
  120. ↑ 120.0 120.1 "Models Overview". https://docs.mistral.ai/getting-started/models/models_overview/. 
  121. ↑ "OLMo 2: The best fully open language model to date | Ai2". https://allenai.org/blog/olmo2. 
  122. ↑ 122.0 122.1 122.2 OLMo, Team; Walsh, Pete; Soldaini, Luca; Groeneveld, Dirk; Lo, Kyle; Arora, Shane; Bhagia, Akshita; Gu, Yuling; Huang, Shengyi (8 October 2025). "2 OLMo 2 Furious". arXiv:2501.00656 [cs.CL].
  123. ↑ "Phi-4 Model Card". 10 July 2025. https://huggingface.co/microsoft/phi-4. 
  124. ↑ "Introducing Phi-4: Microsoft's Newest Small Language Model Specializing in Complex Reasoning". https://techcommunity.microsoft.com/blog/azure-ai-foundry-blog/introducing-phi-4-microsoft%E2%80%99s-newest-small-language-model-specializing-in-comple/4357090. 
  125. ↑ deepseek-ai/DeepSeek-V3, DeepSeek, 26 December 2024, https://github.com/deepseek-ai/DeepSeek-V3?tab=readme-ov-file, retrieved 26 December 2024 
  126. ↑ Feng, Coco (25 March 2025). "DeepSeek wows coders with more powerful open-source V3 model". https://www.scmp.com/tech/big-tech/article/3303798/deepseeks-upgraded-foundational-model-excels-coding-and-maths. 
  127. ↑ Amazon Nova Micro, Lite, and Pro - AWS AI Service Cards3, Amazon, 27 December 2024, https://docs.aws.amazon.com/ai/responsible-ai/nova-micro-lite-pro/overview.html, retrieved 27 December 2024 
  128. ↑ deepseek-ai/DeepSeek-R1, DeepSeek, 21 January 2025, https://github.com/deepseek-ai/DeepSeek-R1, retrieved 21 January 2025 
  129. ↑ DeepSeek-AI; Guo, Daya; Yang, Dejian; Zhang, Haowei; Song, Junxiao; Zhang, Ruoyu; Xu, Runxin; Zhu, Qihao et al. (22 January 2025), "DeepSeek-R1 incentivizes reasoning in LLMS through reinforcement learning", Nature 645 (8081): 633–638, doi:10.1038/s41586-025-09422-z, PMID 40962978, Bibcode: 2025Natur.645..633G 
  130. ↑ Qwen; Yang, An; Yang, Baosong; Zhang, Beichen; Hui, Binyuan; Zheng, Bo; Yu, Bowen; Li, Chengyuan; Liu, Dayiheng (3 January 2025). "Qwen2.5 Technical Report". arXiv:2412.15115 [cs.CL].
  131. ↑ MiniMax-AI/MiniMax-01, MiniMax, 26 January 2025, https://github.com/MiniMax-AI/MiniMax-01?tab=readme-ov-file, retrieved 26 January 2025 
  132. ↑ MiniMax; Li, Aonian; Gong, Bangwei; Yang, Bo; Shan, Boji; Liu, Chang; Zhu, Cheng; Zhang, Chunhao; Guo, Congchao (14 January 2025). "MiniMax-01: Scaling Foundation Models with Lightning Attention". arXiv:2501.08313 [cs.CL].
  133. ↑ Kavukcuoglu, Koray (5 February 2025). "Gemini 2.0 is now available to everyone". https://blog.google/technology/google-deepmind/gemini-model-updates-february-2025/. 
  134. ↑ "Gemini 2.0: Flash, Flash-Lite and Pro". https://developers.googleblog.com/en/gemini-2-family-expands/. 
  135. ↑ Franzen, Carl (5 February 2025). "Google launches Gemini 2.0 Pro, Flash-Lite and connects reasoning model Flash Thinking to YouTube, Maps and Search". VentureBeat. https://venturebeat.com/ai/google-launches-gemini-2-0-pro-flash-lite-and-connects-reasoning-model-flash-thinking-to-youtube-maps-and-search/. 
  136. ↑ "Grok 3 Beta — The Age of Reasoning Agents". https://x.ai/blog/grok-3. 
  137. ↑ "Claude 3.7 Sonnet and Claude Code". https://www.anthropic.com/news/claude-3-7-sonnet. 
  138. ↑ "Introducing GPT-4.5". https://openai.com/index/introducing-gpt-4-5/. 
  139. ↑ Kavukcuoglu, Koray (25 March 2025). "Gemini 2.5: Our most intelligent AI model". https://blog.google/technology/google-deepmind/gemini-model-thinking-updates-march-2025/. 
  140. ↑ "meta-llama/Llama-4-Maverick-17B-128E · Hugging Face". 5 April 2025. https://huggingface.co/meta-llama/Llama-4-Maverick-17B-128E. 
  141. ↑ "The Llama 4 herd: The beginning of a new era of natively multimodal AI innovation". https://ai.meta.com/blog/llama-4-multimodal-intelligence/. 
  142. ↑ "Introducing OpenAI o3 and o4-mini". https://openai.com/index/introducing-o3-and-o4-mini/. 
  143. ↑ Team, Qwen (29 April 2025). "Qwen3: Think Deeper, Act Faster". https://qwenlm.github.io/blog/qwen3/. 
  144. ↑ "Introducing Claude 4". https://www.anthropic.com/news/claude-4. 
  145. ↑ Yadav, Nandini (26 May 2025). "Indian AI startup launches Sarvam-M model: What is it, why is everyone talking about it". https://www.indiatoday.in/technology/features/story/indian-ai-startup-launches-sarvam-m-model-what-is-it-why-is-everyone-talking-about-it-2730778-2025-05-26. 
  146. ↑ "Sarvam-M: Open Source Hybrid Indic LLM | Sarvam AI". 23 May 2025. https://www.sarvam.ai/blogs/sarvam-m. 
  147. ↑ "Grok 4". 9 July 2025. https://x.ai/news/grok-4. 
  148. ↑ Pundalik, Kundeshwar; Sawarkar, Piyush; Sahoo, Nihar; Shinde, Abhishek; Chanda, Prateek; Goswami, Vedant; Nagpal, Ajay; Singh, Atul; Thakur, Viraj (16 July 2025). "PARAM-1 BharatGen 2.9B Model". arXiv:2507.13390 [cs.CL].
  149. ↑ "README.md · bharatgenai/Param-1". 24 February 2026. https://huggingface.co/bharatgenai/Param-1/blob/main/README.md. 
  150. ↑ "zai-org/GLM-4.5 · Hugging Face". 4 August 2025. https://huggingface.co/zai-org/GLM-4.5. 
  151. ↑ "GLM-4.5: Reasoning, Coding, and Agentic Abililties". https://z.ai/blog/glm-4.5. 
  152. ↑ Whitwam, Ryan (5 August 2025). "OpenAI announces two "gpt-oss" open AI models, and you can download them today". https://arstechnica.com/ai/2025/08/openai-releases-its-first-open-source-models-since-2019/. 
  153. ↑ "Claude Opus 4.1". https://www.anthropic.com/news/claude-opus-4-1. 
  154. ↑ "Introducing GPT-5". 7 August 2025. https://openai.com/index/introducing-gpt-5/. 
  155. ↑ "OpenAI Platform: GPT-5 Model Documentation". https://platform.openai.com/docs/models/gpt-5. 
  156. ↑ "deepseek-ai/DeepSeek-V3.1 · Hugging Face". 21 August 2025. https://huggingface.co/deepseek-ai/DeepSeek-V3.1. 
  157. ↑ "DeepSeek-V3.1 Release | DeepSeek API Docs". https://api-docs.deepseek.com/news/news250821. 
  158. ↑ Kirchner, Malte (2 September 2025). "Apertus: Schweiz stellt erstes offenes und mehrsprachiges KI-Modell vor" (in de). heise online. https://www.heise.de/news/Apertus-Schweiz-stellt-erstes-offenes-und-mehrsprachiges-KI-Modell-vor-10629412.html. 
  159. ↑ "Apertus: Ein vollständig offenes, transparentes und mehrsprachiges Sprachmodell" (in de). Zürich: ETH Zürich. 2 September 2025. https://ethz.ch/de/news-und-veranstaltungen/eth-news/news/2025/09/medienmitteilung-apertus-ein-vollstaendig-offenes-transparentes-und-mehrsprachiges-sprachmodell.html. 
  160. ↑ "Introducing Claude Sonnet 4.5". https://www.anthropic.com/news/claude-sonnet-4-5. 
  161. ↑ "GLM-4.6: Advanced Agentic, Reasoning and Coding Capabilities". https://z.ai/blog/glm-4.6. 
  162. ↑ "zai-org/GLM-4.6 · Hugging Face". 30 September 2025. https://huggingface.co/zai-org/GLM-4.6. 
  163. ↑ "GLM-4.6". https://modelscope.cn/models/ZhipuAI/GLM-4.6. 
  164. ↑ "A new era of intelligence with Gemini 3". 18 November 2025. https://blog.google/products/gemini/gemini-3/. 
  165. ↑ "Olmo 3: Charting a path through the model flow to lead open-source AI". 20 November 2025. https://allenai.org/blog/olmo3. 
  166. ↑ Olmo, Team; Ettinger, Allyson; Bertsch, Amanda; Kuehl, Bailey; Graham, David; Heineman, David; Groeneveld, Dirk; Brahman, Faeze; Timbers, Finbarr (15 December 2025). "Olmo 3". arXiv:2512.13961 [cs.CL].
  167. ↑ "Introducing Claude Opus 4.5". https://www.anthropic.com/news/claude-opus-4-5. 
  168. ↑ Binder, Matt (3 December 2025). "DeepSeek v3.2: What it is, how it compares to ChatGPT, how to try it". https://mashable.com/article/deepseek-v3-2-models-released. 
  169. ↑ "DeepSeek-V3.2 Release". 1 December 2025. https://api-docs.deepseek.com/news/news251201. 
  170. ↑ "DeepSeek-V3.2: Efficient Reasoning & Agentic AI". 1 December 2025. https://huggingface.co/deepseek-ai/DeepSeek-V3.2. 
  171. ↑ "Advancing science and math with GPT-5.2". https://openai.com/index/gpt-5-2-for-science-and-math/. 
  172. ↑ "Pushing Qwen3-Max-Thinking Beyond its Limits". 25 January 2026. https://qwen.ai/blog?id=qwen3-max-thinking. "We further enhance Qwen3-Max-Thinking with two key innovations: (1) adaptive tool-use capabilities [...]; and (2) advanced test-time scaling techniques [...]. [...] We limit [parallel trajectories] and redirect saved computation to iterative self-reflection guided by a “take-experience” mechanism." 
  173. ↑ "K2: An open source model that delivers frontier capabilities". Mohamed bin Zayed University of Artificial Intelligence. 5 December 2025. https://mbzuai.ac.ae/news/k2-an-open-source-model-that-delivers-frontier-capabilities. 
  174. ↑ "K2 Think V2: a fully sovereign reasoning model". Mohamed bin Zayed University of Artificial Intelligence. 27 January 2026. https://mbzuai.ac.ae/news/k2-think-v2-a-fully-sovereign-reasoning-model. 
  175. ↑ "LICENSE · moonshotai/Kimi-K2.5". 1 January 2026. https://huggingface.co/moonshotai/Kimi-K2.5/blob/70a8ad02d86dbb7459abd22b8a4feca294d2c993/LICENSE. 
  176. ↑ Team, Kimi; Bai, Yifan; Bao, Yiping; Charles, Y.; Chen, Cheng; Chen, Guanduo; Chen, Haiting; Chen, Huarong; Chen, Jiahao (3 February 2026). "Kimi K2: Open Agentic Intelligence". arXiv:2507.20534 [cs.LG].
  177. ↑ Team, Kimi; Bai, Tongtong; Bai, Yifan; Bao, Yiping; Cai, S. H.; Cao, Yuan; Charles, Y.; Che, H. S.; Chen, Cheng (2 February 2026). "Kimi K2.5: Visual Agentic Intelligence". arXiv:2602.02276 [cs.CL].
  178. ↑ "Kimi K2.5: Chat with Kimi K2.5 for Free". https://kimi-k25.com/blog/kimi-k2-5-agent-swarm. 
  179. ↑ Jiang, Ben (3 February 2026). "Compact AI model from China's StepFun outshines rivals from DeepSeek, Moonshot". https://www.scmp.com/tech/article/3342222/punches-above-its-weight-compact-ai-model-chinas-stepfun-outshines-larger-rivals. 
  180. ↑ "Step 3.5 Flash: Fast Enough to Think. Reliable Enough to Act.". 12 February 2026. https://static.stepfun.com/blog/step-3.5-flash/. 
  181. ↑ "stepfun-ai/Step-3.5-Flash". 14 March 2026. https://huggingface.co/stepfun-ai/Step-3.5-Flash. 
  182. ↑ Mo, Liam; Goh, Brenda; Chen, Laurie (11 February 2026). "Chinese AI startup Zhipu releases new flagship model GLM-5". Reuters. https://www.reuters.com/technology/chinas-ai-startup-zhipu-releases-new-flagship-model-glm-5-2026-02-11/. 
  183. ↑ 183.0 183.1 "GLM-5 / README.md". 27 August 2026. https://github.com/zai-org/GLM-5/blob/c9d779052c2fa214b09d36a75c58c78ca9509578/README.md. 
  184. ↑ 184.0 184.1 Butts, Dylan (16 February 2026). "Alibaba unveils Qwen3.5 as China's chatbot race shifts to AI agents". https://www.cnbc.com/2026/02/17/china-alibaba-qwen-ai-agent-latest-model.html. 
  185. ↑ 185.0 185.1 "Qwen3.5: Towards Native Multimodal Agents". 15 February 2026. https://qwen.ai/blog?id=qwen3.5. 
  186. ↑ "LICENSE · Qwen/Qwen3.5-397B-A17B". 16 February 2026. https://huggingface.co/Qwen/Qwen3.5-397B-A17B/blob/7f74e137107f2d2e056d995ef37f898a1f270d8d/LICENSE. 
  187. ↑ "LICENSE · bharatgenai/Param2-17B-A2.4B-Thinking". 16 February 2026. https://huggingface.co/bharatgenai/Param2-17B-A2.4B-Thinking/blob/9c8cc097a1211f1f18d177ab99e73846ca549c38/LICENSE. 
  188. ↑ "bharatgenai/Param2-17B-A2.4B-Thinking". https://huggingface.co/bharatgenai/Param2-17B-A2.4B-Thinking. 
  189. ↑ "sarvamai/sarvam-105b · Hugging Face". https://huggingface.co/sarvamai/sarvam-105b. 
  190. ↑ Kumar, Abhijeet (19 February 2026). "Why Sarvam's new 105B model marks a shift in India's sovereign AI ambitions". https://www.business-standard.com/technology/tech-news/sarvam-105b-model-sovereign-ai-india-foundation-model-launch-impact-summit-126021900551_1.html. 
  191. ↑ Singh, Jagmeet (18 February 2026). "Indian AI lab Sarvam's new models are a major bet on the viability of open source AI" (in en-US). https://techcrunch.com/2026/02/18/indian-ai-lab-sarvams-new-models-are-a-major-bet-on-the-viability-of-open-source-ai/. 
  192. ↑ "Open-Sourcing Sarvam 30B and 105B". 6 March 2026. https://www.sarvam.ai/blogs/sarvam-30b-105b. 
  193. ↑ "sarvamai/sarvam-1-v0.5 · Hugging Face". https://huggingface.co/sarvamai/sarvam-1-v0.5. 
  194. ↑ Marquez, Javier (17 March 2026). "Una IA para reunir todas las funciones posibles: la apuesta de Mistral con Small 4 es hacer más con menos cosas" (in es). https://www.xataka.com/robotica-e-ia/europea-mistral-acaba-lanzar-small-4-su-apuesta-carrera-ia-reunir-varias-funciones-solo-modelo. 
  195. ↑ "Introducing Mistral Small 4". https://mistral.ai/news/mistral-small-4. 
  196. ↑ "Mystery AI model revealed to be Xiaomi's following suspicions it was DeepSeek's". Reuters. 18 March 2026. https://www.reuters.com/business/media-telecom/mystery-ai-model-has-developers-buzzing-is-this-deepseeks-latest-blockbuster-2026-03-18/. 
  197. ↑ "Xiaomi Launches Powerful AI Model MiMo-V2 Pro With 1 Trillion Parametres, 1 Million Token Context Window". NDTV Profit. 19 March 2026. https://www.ndtvprofit.com/technology/xiaomi-launches-powerful-ai-model-mimo-v2-pro-with-1-trillion-parametres-1-million-token-context-window-11236705. 
  198. ↑ "Qwen 3.6-Plus has been released, featuring an agent function that autonomously performs tasks.". Gigazine. 3 April 2026. https://gigazine.net/gsc_news/en/20260403-qwen-3-6-plus/. 
  199. ↑ "Qwen3.6-Plus: Towards Real World Agents". 1 April 2026. https://qwen.ai/blog?id=qwen3.6. 
  200. ↑ Whitwam, Ryan (2 April 2026). "Google announces Gemma 4 open AI models, switches to Apache 2.0 license". https://arstechnica.com/ai/2026/04/google-announces-gemma-4-open-ai-models-switches-to-apache-2-0-license/. 
  201. ↑ Mann, Tobias (2 April 2026). "Google battles Chinese open weights models with Gemma 4". https://www.theregister.com/2026/04/02/googles_gemma_4_open_weights/. 
  202. ↑ Franzen, Carl (7 April 2026). "AI joins the 8-hour work day as GLM ships 5.1 open source LLM, beating Opus 4.6 and GPT-5.4 on SWE-Bench Pro". https://venturebeat.com/technology/ai-joins-the-8-hour-work-day-as-glm-ships-5-1-open-source-llm-beating-opus-4. 
  203. ↑ "GLM-5.1: Towards Long-Horizon Tasks". https://z.ai/blog/glm-5.1. 
  204. ↑ "Introducing Muse Spark: Scaling Towards Personal Superintelligence". 8 April 2026. https://ai.meta.com/blog/introducing-muse-spark-msl/. 
  205. ↑ "A Chinese AI called 'Qwen3.6-35B-A3B,' which is more powerful than Gemma4, has been released as an open model.". 17 April 2026. https://gigazine.net/gsc_news/en/20260417-qwen36-35b-a3b. 
  206. ↑ "README.md · Qwen/Qwen3.6-35B-A3B". 15 April 2026. https://huggingface.co/Qwen/Qwen3.6-35B-A3B/blob/main/README.md. 
  207. ↑ "LICENSE · moonshotai/Kimi-K2.6". 14 April 2026. https://huggingface.co/moonshotai/Kimi-K2.6/blob/5a49d036ab7472b7d5912ded487150ec1358c11d/LICENSE. 
  208. ↑ Gewirtz, David (21 April 2026). "Moonshot AI's new Kimi K2.6 swarms your complex tasks with 1,000 collaborating agents". https://www.zdnet.com/article/moonshot-ai-kimi-k2-6-swarms-complex-tasks-collaborating-agents/. 
  209. ↑ Li, Yutong (22 April 2026). "Error: no |title= specified when using {{Cite web}}". in Du, Yu (in zh). https://www.nbd.com.cn/articles/2026-04-22/4351579.html. 
  210. ↑ "README.md · moonshotai/Kimi-K2.6". 23 April 2026. https://huggingface.co/moonshotai/Kimi-K2.6/blob/5a49d036ab7472b7d5912ded487150ec1358c11d/README.md. 
  211. ↑ "LICENSE · Tencent-Hunyuan/Hy3-preview". https://github.com/Tencent-Hunyuan/Hy3-preview/blob/main/LICENSE. 
  212. ↑ "Tencent Unveils Hy3 preview; Model Enhances Agent Capabilities and Real-World Usability". 24 April 2026. https://www.tencent.com/index.php/en-us/articles/2202320.html. 
  213. ↑ 213.0 213.1 "Chinese AI startup DeepSeek quietly released 'DeepSeek V4 Pro 0813'.". 13 August 2026. https://gigazine.net/gsc_news/en/20260813-deepseek-v4-pro-0813. 
  214. ↑ "DeepSeek Doubles Down on Low-Cost AI With V4-Flash Update". 3 August 2026. https://www.mitsloanme.com/article/deepseek-doubles-down-on-low-cost-ai-with-v4-flash-update/. 
  215. ↑ Butts, Dylan (24 April 2026). "China's DeepSeek releases preview of long-awaited V4 model as AI race intensifies". https://www.cnbc.com/2026/04/24/deepseek-v4-llm-preview-open-source-ai-competition-china.html. 
  216. ↑ Baptista, Eduardo (3 August 2026). "DeepSeek's new AI model is by far the cheapest of well-known models to run, research firm says". https://www.reuters.com/business/retail-consumer/deepseeks-new-ai-model-is-by-far-cheapest-well-known-models-run-research-firm-2026-08-03/. 
  217. ↑ "MiMo-V2.5-Pro | Xiaomi". https://mimo.xiaomi.com/mimo-v2-5-pro. 
  218. ↑ Thomas, Prasanth Aby (28 April 2026). "Xiaomi releases MIT-licensed MiMo models for long-running AI agents". https://www.computerworld.com/article/4164220/xiaomi-releases-mit%E2%80%91licensed-mimo-models-for-long%E2%80%91running-ai-agents-2.html. 
  219. ↑ "XiaomiMiMo/MiMo-V2.5". XiaomiMiMo. 27 April 2026. https://huggingface.co/XiaomiMiMo/MiMo-V2.5. 
  220. ↑ "LICENSE · mistralai/Mistral-Medium-3.5-128B". 29 April 2026. https://huggingface.co/mistralai/Mistral-Medium-3.5-128B/blob/32f566da413d9fde169ee1276a7551f4a2d660fb/LICENSE. 
  221. ↑ "Mistral Medium 3.5 - Mistral AI". https://docs.mistral.ai/models/model-cards/mistral-medium-3-5-26-04. 
  222. ↑ "Composer 2.5". 18 May 2026. https://cursor.com/changelog/composer-2-5. 
  223. ↑ "Gemini 3.5: frontier intelligence with action" (in en-us). 19 May 2026. https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-5/. 
  224. ↑ Bhati, Divya (28 May 2026). "Alibaba AI Qwen latest model beats Google Gemini and ChatGPT in coding". India Today. https://www.indiatoday.in/technology/news/story/alibaba-ai-qwen-latest-model-beats-google-gemini-and-chatgpt-in-coding-2918356-2026-05-28. 
  225. ↑ "Qwen3.7: The Agent Frontier". 20 May 2026. https://qwen.ai/blog?id=qwen3.7. 
  226. ↑ "Introducing Claude Opus 4.8". 28 May 2026. https://www.anthropic.com/news/claude-opus-4-8. 
  227. ↑ "Step 3.7 Flash". 29 May 2026. https://static.stepfun.com/blog/step-3.7-flash/. 
  228. ↑ "LICENSE · MiniMaxAI/MiniMax-M3". 12 June 2026. https://huggingface.co/MiniMaxAI/MiniMax-M3/blob/3a41b311ffa5719cef48fed3974ccf2cc03733ea/LICENSE. 
  229. ↑ Chang, Minxiao (1 June 2026). "MiniMax debuts AI model built for long and complex coding tasks". South China Morning Post. https://www.scmp.com/tech/tech-trends/article/3355529/minimax-debuts-ai-model-built-long-and-complex-coding-tasks. 
  230. ↑ "README.md · MiniMaxAI/MiniMax-M3". 12 June 2026. https://huggingface.co/MiniMaxAI/MiniMax-M3/blob/3a41b311ffa5719cef48fed3974ccf2cc03733ea/README.md. 
  231. ↑ 231.0 231.1 "OpenMDW-1.1". https://openmdw.ai/license/1-1/. 
  232. ↑ "NVIDIA announces its most powerful open-source model in the US, the 'Nemotron 3 Ultra,' and reports the start of mass production of its AI server, 'Vera Rubin.'". Gigazine. 1 June 2026. https://gigazine.net/gsc_news/en/20260601-nemotron-3-ultra/. 
  233. ↑ "README.md · nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-NVFP4". 24 June 2026. https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-NVFP4/blob/183968f87ae4cedce3039313cac1fd43d112c578/README.md. 
  234. ↑ "Claude Fable 5 and Claude Mythos 5". 9 June 2026. https://www.anthropic.com/news/claude-fable-5-mythos-5. 
  235. ↑ "LICENSE · moonshotai/Kimi-K2.7-Code". 11 June 2026. https://huggingface.co/moonshotai/Kimi-K2.7-Code/blob/da26fe4d239c5c9d14507f1f909894201f911ede/LICENSE. 
  236. ↑ "The Chinese-made AI model 'Kimi K2.7 Code' has been released as an open model, boasting the best coding performance in the Kimi series and low token consumption.". 15 June 2026. https://gigazine.net/gsc_news/en/20260615-kimi-k2-7-code/. 
  237. ↑ "README.md · moonshotai/Kimi-K2.7-Code". 12 June 2026. https://huggingface.co/moonshotai/Kimi-K2.7-Code/blob/78b5c461e9cd0ffa168d8b5c2151bcf6b0799d7d/README.md. 
  238. ↑ Chang, Minxiao (15 June 2026). "Zhipu AI's stock rockets after Chinese firm launches GLM-5.2 model". https://www.scmp.com/tech/tech-trends/article/3357115/zhipu-ais-stock-rockets-after-chinese-firm-makes-glm-52-open-source. 
  239. ↑ "China's Z.ai GLM-5.2 tops OpenAI's GPT 5.5 model on key benchmarks". 17 June 2026. https://economictimes.indiatimes.com/tech/artificial-intelligence/chinas-z-ai-glm-5-2-tops-openais-gpt-5-5-model-on-key-benchmarks/articleshow/131805202.cms. 
  240. ↑ "GLM-5.2: Built for Long-Horizon Tasks". https://z.ai/blog/glm-5.2. 
  241. ↑ Franzen, Carl (29 June 2026). "Meituan open sources LongCat-2.0, the 1.6T, near-frontier agentic coding model that's been leading OpenRouter — trained entirely on Chinese chips". https://venturebeat.com/technology/meituan-open-sources-longcat-2-0-the-1-6t-near-frontier-agentic-coding-model-thats-been-leading-openrouter-trained-entirely-on-chinese-chips. 
  242. ↑ "Introducing LongCat-2.0". 30 June 2026. https://longcat.chat/blog/longcat-2.0/. 
  243. ↑ Witteveen, Sam (6 July 2026). "Tencent's Apache-licensed Hy3 takes on GLM-5.2 at half the size — and wins everywhere except coding". VentureBeat. https://venturebeat.com/technology/tencents-apache-licensed-hy3-takes-on-glm-5-2-at-half-the-size-and-wins-everywhere-except-coding. 
  244. ↑ "Introducing Grok 4.5". https://x.ai/news/grok-4-5. 
  245. ↑ "SpaceXAI releases Grok 4.5, which Elon describes as an 'Opus-class model'". 8 July 2026. https://techcrunch.com/2026/07/08/spacexai-releases-grok-4-5-which-elon-describes-as-an-opus-class-model/. 
  246. ↑ "SWE-1.7". 8 July 2026. https://cognition.com/blog/swe-1-7. 
  247. ↑ Varghese, Harshita (9 July 2026). "Meta debuts Muse Spark 1.1 model with preview open to developers". Reuters. https://www.reuters.com/business/meta-debuts-muse-spark-11-with-preview-open-developers-2026-07-09/. 
  248. ↑ Loizos, Connie (15 July 2026). "Thinking Machines amps up its bet against one-size-fits-all AI with its first open model, Inkling". https://techcrunch.com/2026/07/15/thinking-machines-amps-up-its-bet-against-one-size-fits-all-ai-with-its-first-open-model-inkling/. 
  249. ↑ Franzen, Carl (15 July 2026). "Thinking Machines open sources first multimodal language model, Inkling, focused on low cost and 'resistance to censorship'". https://venturebeat.com/technology/thinking-machines-open-sources-first-multimodal-language-model-inkling-focused-on-low-cost-and-resistance-to-censorship. 
  250. ↑ Franzen, Carl (27 July 2026). "Kimi K3's full weights are here, but they're 'open' with a caveat: What enterprises should know". https://venturebeat.com/technology/kimi-k3s-full-weights-are-here-but-theyre-open-with-a-caveat-what-enterprises-should-know. 
  251. ↑ "Moonshot AI releases Kimi K3 model weights under custom licence: What it means for enterprises". The Indian Express. 28 July 2026. https://indianexpress.com/article/technology/artificial-intelligence/china-moonshot-ai-kimi-k3-model-weights-custom-licence-10807478/. 
  252. ↑ Yildirim, Ece (16 July 2026). "China Just Dropped Another Bomb on America's Frontier AI Companies". Gizmodo. https://gizmodo.com/china-just-dropped-another-bomb-on-americas-frontier-ai-companies-2000786670. 
  253. ↑ 253.0 253.1 "Qwen3.8, a Chinese AI with 2.4 trillion parameters, has been released as an open model for free, boasting performance comparable to the most advanced models from OpenAI and Anthropic.". 13 August 2026. https://gigazine.net/gsc_news/en/20260813-alibaba-qwen-3-8-open/. 
  254. ↑ 254.0 254.1 "Error: no |title= specified when using {{Cite web}}". 14 August 2026. https://www.ithome.com/0/989/953.htm. 
  255. ↑ "Alibaba's Qwen Unveils Preview of Flagship AI Model". Bloomberg News. 19 July 2026. https://www.bloomberg.com/news/articles/2026-07-19/alibaba-s-qwen-unveils-preview-of-flagship-ai-model. 
  256. ↑ 256.0 256.1 "Introducing Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber" (in en-us). 21 July 2026. https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-6-flash-3-5-flash-lite-3-5-flash-cyber/. 
  257. ↑ Martin, Iain (21 July 2026). "American Open-Source Labs Think They Can Beat China's Best AI Startups". https://www.forbes.com/sites/iainmartin/2026/07/21/american-open-source-labs-think-they-can-beat-chinas-best-ai-startups/. 
  258. ↑ Nuñez, Michael (21 July 2026). "Poolside drops Laguna S 2.1, an open-weight coding model that beats rivals 10x its size". https://venturebeat.com/infrastructure/poolside-drops-laguna-s-2-1-an-open-weight-coding-model-that-beats-rivals-10x-its-size. 
  259. ↑ Yoon, Jeong-hoon (21 July 2026). "Error: no |title= specified when using {{Cite web}}" (in ko). https://www3.edaily.co.kr/News/Read?newsId=03460406645516816. 
  260. ↑ Kwon, Yong-man (13 August 2026). "Error: no |title= specified when using {{Cite web}}". https://it.chosun.com/news/articleView.html?idxno=2023092168121. 
  261. ↑ "README.md · Motif-Technologies/Motif-3". 12 August 2026. https://huggingface.co/Motif-Technologies/Motif-3/blob/ccceb1a5fd7b5eb32e47841216b3caf5666c07bc/README.md. 
  262. ↑ "LICENSE · upstage/Solar-Open2-250B". 22 July 2026. https://huggingface.co/upstage/Solar-Open2-250B/blob/8d1dda5ad2af58207c65348d825772d2f4d72481/LICENSE. 
  263. ↑ Jhoo, Dong-chan (23 July 2026). "Upstage unveils Solar Open 2, open-weight AI model built for autonomous tasks". https://www.koreatimes.co.kr/business/companies/20260723/upstage-unveils-solar-open-2-open-weight-ai-model-built-for-autonomous-tasks. 
  264. ↑ "README.md · upstage/Solar-Open2-250B". 31 July 2026. https://huggingface.co/upstage/Solar-Open2-250B/blob/9190fbe63a2ad8e17fc766ccceb36de7c66f004b/README.md. 
  265. ↑ "Fugu". https://sakana.ai/fugu. 
  266. ↑ Brandom, Russell (24 July 2026). "Anthropic launches Opus 5". https://techcrunch.com/2026/07/24/anthropic-launches-opus-5/. 
  267. ↑ Shim, Min-gwan (29 July 2026). "SK Telecom debuts A.X K2, drives Korea's AI into industry, defense, biotech". https://biz.chosun.com/en/en-it/2026/07/29/X6RUYHAE25BWREQ2XMRA7IBJLA/. 
  268. ↑ "README.md · skt/A.X-K2". 8 August 2026. https://huggingface.co/skt/A.X-K2/blob/7b3969c2a84b43c03718b9acc82e15a68b1a4c5a/README.md. 
  269. ↑ Franzen, Carl (30 July 2026). "Thinking Machines debuts Inkling Small open source AI model nearing performance of predecessor at about 1/4 size". https://venturebeat.com/technology/thinking-machines-debuts-inkling-small-open-source-ai-model-nearing-performance-of-predecessor-at-about-1-4-size. 
  270. ↑ Jo, He-rim (31 July 2026). "LG unveils Korea's largest AI model with 750 billion parameters". https://www.koreaherald.com/article/10827054. 
  271. ↑ Choi, A-ri (10 August 2026). "Four AI Teams Face Elimination in National Project Evaluation". https://www.chosun.com/english/industry-en/2026/08/10/AQ2U237LPRGOJGTQJZB2TSZSGE/. 
  272. ↑ Mohanty, Tamanna (10 August 2026). "Meta launches Muse Glimmer, a 30B open-weight AI model designed to run locally". Fortune India. https://www.fortuneindia.com/technology/meta-launches-muse-glimmer-a-30b-open-weight-ai-model-designed-to-run-locally/153062. 
  273. ↑ "Gemini 3.7 Flash" (in en-us). 13 August 2026. https://deepmind.google/models/model-cards/gemini-3-7-flash/. 
  274. ↑ "LICENSE · tencent/Hy4-preview". 28 August 2026. https://huggingface.co/tencent/Hy4-preview/blob/79c38554846799f70bcd54b9142eac71fbecafaf/LICENSE. 
  275. ↑ Bhati, Divya (28 August 2026). "Tencent releases new AI model, says it beat ZAI and Moonshoot in testing". https://www.indiatoday.in/technology/news/story/tencent-releases-new-ai-model-says-it-beat-zai-and-moonshoot-in-testing-2981956-2026-08-28. 
  276. ↑ "Tencent Touts New AI Model It Claims Outperforms Z.AI, Moonshot". 28 August 2026. https://www.bloomberg.com/news/articles/2026-08-28/tencent-touts-new-ai-model-it-claims-outperforms-z-ai-moonshot. 
  277. ↑ 277.0 277.1 Patterson, Ben (1 September 2026). "Claude's updated Fable and Mythos models watermark their replies". https://www.pcworld.com/article/3224834/claudes-updated-fable-and-mythos-models-watermark-their-replies.html. 
  278. ↑ "Gemini 3.8 Flash" (in en-us). 2 September 2026. https://deepmind.google/models/model-cards/gemini-3-8-flash/. 
  279. ↑ Bensinger, Greg (3 September 2026). "OpenAI launches new Astra model amid growing scrutiny over agents' safety". Reuters (San Francisco). https://www.reuters.com/legal/litigation/openai-launches-new-astra-model-amid-growing-scrutiny-over-agents-safety-2026-09-03/. 
  280. ↑ "Introducing K2 Horizon: Frontier Performance, Radically Open". Mohamed bin Zayed University of Artificial Intelligence. 3 September 2026. https://ifm.ai/blog/k2. 
  281. ↑ Haralu, Kileuna (10 September 2026). "DeepSeek Launches V4.1-Flash to Make Large-Scale AI Inference Cheaper". Analytics India Mag. https://analyticsindiamag.com/ai-news/deepseek-launches-v41-flash-to-make-large-scale-ai-inference-cheaper. 
  282. ↑ "README.md · deepseek-ai/DeepSeek-V4.1-Flash". 10 September 2026. https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash/blob/df42c109f1defefcbfcedbe7d905718a12266e40/README.md. 
  283. ↑ "Introducing Grok 4.7". SpaceXAI. September 21, 2026. https://x.ai/news/grok-4-7. 
  284. ↑ Wright, Webb (22 September 2026). "OpenAI Releases New GPT-6 Sol and Luna Models, Sells Them as Mini-Astras". Gizmodo. https://gizmodo.com/openai-releases-new-gpt-6-sol-and-luna-models-sells-them-as-mini-astras-2000815731. 
  285. ↑ 285.0 285.1 285.2 285.3 "MiMo-V2.6: Scaling Reinforcement Learning Toward Self-Improvement". XiaomiMiMo. 22 September 2026. https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Pro-RL/blob/main/MiMo_V2_6_technical_report.pdf. 
  286. ↑ "Introducing MiMo-V2.6 series". 22 September 2026. https://mimo.xiaomi.com/mimo-v2-6. 
  287. ↑ "Anthropic unveils Claude Opus 5.5". Reuters. 22 September 2026. https://www.reuters.com/business/anthropic-unveils-claude-opus-55-2026-09-22/. 
  288. ↑ Capoot, Ashley (28 September 2026). "Anthropic launches cheaper AI model, its second release since CEO’s call for a slowdown". https://www.cnbc.com/2026/09/28/anthropic-sonnet-5-5-launch.html. 
  289. ↑ Dellinger, Aj (29 September 2026). "With No Astra to Release, OpenAI Pivots to New GPT-6.1 Sol Model". https://gizmodo.com/with-no-astra-to-release-openai-pivots-to-new-gpt-6-1-sol-model-2000819044. 
  290. ↑ Subin, Samantha; Sigalos, MacKenzie. "Google rolls out Gemini 4 Argon, its most advanced AI model". https://www.cnbc.com/2026/09/30/google-gemini-4-argon-ai.html. 

Template:Large language models



Retrieved from "https://handwiki.org/wiki/index.php?title=Software:List_of_large_language_models&oldid=4627383"