Comparison of generative AI models

From HandWiki
Short description: none


This is a comparison of generative AI models, such as language models and text-to-image models, across various benchmarks and user-generated Elo ratings. Leading large language models are sometimes called frontier models.

Background

Independent evaluation is widely used to compare frontier models because developer-reported scores are hard to compare directly.

  • Artificial Analysis (AA) runs the Artificial Analysis Intelligence Index v4.3.2 across 10 evaluations including AA-Briefcase, GDPval-AA, AutomationBench-AA and Terminal-Bench 4.0, plus price per 1M tokens and output speed in tokens/s.[1][2] Its pricing and index data have been cited in coverage reporting DeepSeek V4-Flash as by far the cheapest well-known model to run.[3]
  • Arena.ai, formerly Chatbot Arena from University of California, Berkeley, ranks models by blind human votes into elo-style ratings.[4][5] It tripled to a $1.7 billion valuation in January 2026 after raising $150 million.[4] Earlier coverage described its spin-out from an academic project into a $600 million startup.[6]
  • Epoch AI publishes the Epoch Capabilities Index (ECI), a composite across 50 plus benchmarks scaled so that Claude 3.5 Sonnet equals 130.[7] GPT-6 Astra set a new record at 166 in September 2026, ahead of Claude Fable 5.1 at 164.[8] Its independence has been scrutinized over delayed disclosure of OpenAI funding for FrontierMath.[9]
  • Vals AI publishes the Vals Index, agentic performance across finance, coding and legal tasks weighted by share of United States GDP, updated 10 September 2026 with 56 models tested.[10]

Language models

Comparison of language models across benchmarks, as of 22 September 2026[lower-alpha 1]
Model Developer Origin Release Date License AA Intelligence Index[1][lower-alpha 2] Arena.ai Text Rating[11][lower-alpha 3] Epoch Capability Index[12] Vals Index[13] AA Cost per Task (USD)[1] AA Speed (Median Tokens/s)[1]
Claude Fable 5.1 Anthropic  United States September 2026 Proprietary[14] 53 1498 164 68.8 $7.63 66
GPT-6 Astra OpenAI  United States 3 September 2026 Proprietary[15] 53 1479 166 66.6 $3.26 67
Claude Opus 5 Anthropic  United States 24 July 2026 Proprietary[14] 51 1492 162 67.2 $5.86 56
Claude Fable 5 Anthropic  United States 9 June 2026 Proprietary[14] 50[16] 1506 163 66.0 $8.75[17] 68
Muse Spark 1.3 Meta Platforms  United States September 2026 Proprietary[18] 48 1493 n/a 64.5 $1.60 223
GPT-5.6 Sol OpenAI  United States 9 July 2026 Proprietary[15] 47 1483 161 63.7 $1.99 79
Grok 4.7 xAI  United States September 2026 Proprietary[19] 46 n/a n/a n/a $3.74 39
MiMo-V2.6-Pro Xiaomi  China 21 September 2026 MIT License[20] 46 n/a n/a n/a $0.13 130
Qwen3.8 Max 0902 Alibaba Cloud  China 2 September 2026 Proprietary[21] 45 1481 156 51.8 $5.41 39
GLM-5.3 Zhipu AI  China August 2026 MIT License[22] 45 n/a 155 57.0 $2.01 57
Grok 4.6 xAI  United States 12 August 2026 Proprietary[19] 44 1456 156 59.2 $1.86 60
Step 5 Preview StepFun  China August 2026 Proprietary[23] 44 n/a n/a n/a $0.71 92
Kimi K3 Moonshot AI  China 16 July 2026 Kimi K3 License[24] 44 1485 157 57.8 $2.00 43
GPT-5.6 Terra OpenAI  United States 9 July 2026 Proprietary[15] 42 1466 159 59.6 $1.40 107
GLM-5.3-Flash Zhipu AI  China August 2026 MIT License[22] 42 n/a 151 47.2 $0.25 89
Claude Opus 4.8 Anthropic  United States 28 May 2026 Proprietary[14] 42 n/a 158 n/a $4.08 57
Gemini 3.8 Flash Google DeepMind  United States September 2026 Proprietary[25] 41 1493 156 62.3 $1.24 329
Claude Opus 4.7 Anthropic  United States 16 April 2026 Proprietary[14] 41 1502 156 n/a n/a 50
Qwen3.8 2.4T A95B Alibaba Cloud  China September 2026 Qwen3.8-Max license[21] 40 n/a 156 n/a $2.16 39
Qwen3.8-Flash-Next Alibaba Cloud  China September 2026 Qwen Community 1.0[21] 40 n/a n/a n/a $0.37 54
Gemini 3.7 Flash Google DeepMind  United States 13 August 2026 Proprietary[25] 39 1490 157 59.3 $0.93 318
DeepSeek V4.1 Flash DeepSeek  China August 2026 MIT License[26] 39 n/a n/a 57.9 $0.27 219
GPT-5.4 OpenAI  United States 5 March 2026 Proprietary[15] 39 1476 156 n/a n/a 142
Grok 4.5 xAI  United States 8 July 2026 Proprietary[19] 39 1468 153 n/a $1.04 61
Claude Sonnet 5 Anthropic  United States 30 June 2026 Proprietary[14] 38 1461 156 59.6 $5.09 77
GPT-5.5 OpenAI  United States 23 April 2026 Proprietary[15] 38 1482 159 57.4 $2.63 108
GPT-5.6 Luna OpenAI  United States 9 July 2026 Proprietary[15] 37 1452 156 59.9 $0.18 159
DeepSeek V4 Pro 0813 DeepSeek  China 13 August 2026 MIT License[26] 36 1463 155 52.4 $0.67 97
Agnes 3.0 Flash Sapiens AI  Singapore 2026 Proprietary[27] 36* n/a n/a n/a n/a n/a
Agnes 2.5 Pro Beta Sapiens AI  Singapore 2026 Proprietary[27] 35* n/a n/a n/a n/a n/a
DeepSeek V4 Flash Vision DeepSeek  China August 2026 MIT License[26] 35 n/a n/a n/a $0.31 235
Qwen3.8 27B Alibaba Cloud  China September 2026 Apache 2.0[28] 34 n/a n/a 48.5 $0.82 45
GLM-5.2 Zhipu AI  China 16 June 2026 MIT License[22] 34 n/a 151 53.1 $0.96 71
Gemini 3.6 Flash Google DeepMind  United States 21 July 2026 Proprietary[25] 34 1480 154 n/a $0.93 209
Motif 3 Motif Technologies  South Korea 14 July 2026 MIT License[29] 34* n/a n/a n/a n/a n/a
GPT-5.3 Codex OpenAI  United States 5 February 2026 Proprietary[15] 33* n/a 156 n/a n/a 170
K2 Horizon 375B A23B Institute of Foundation Models  United Arab Emirates August 2026 Apache 2.0[30] 31 n/a n/a n/a n/a n/a
Apodex 1.1 Apodex  United States[31][32][33] August 2026 Proprietary[34] 30* n/a n/a n/a n/a n/a
Gemini 3.1 Pro Preview Google DeepMind  United States 19 February 2026 Proprietary[25] 30 1487 155 41.9 $0.67 115
Qwen3.7 Max Alibaba Cloud  China 19 May 2026 Proprietary[35] 29 1473 n/a n/a $1.15 197

Text-to-image models

Comparison of text-to-image models, as of 20 September 2026[lower-alpha 4]
Model Developer Origin Release Date AA Elo[36][lower-alpha 2] Arena Elo[37] API Price per 1k Images (USD)[36] License
GPT Image 2.5 Sunburst OpenAI  United States September 2026 1182 1421[lower-alpha 4] $210.70 Proprietary[36]
GPT Image 2.5 Flare OpenAI  United States September 2026 1188 1399[lower-alpha 4] $210.70 Proprietary[36]
GPT Image 2 OpenAI  United States April 2026 1171 1381 $211.00 Proprietary[36]
MAI-Image-2.6 Microsoft AI  United States August 2026 1147 1331 $38.90 Proprietary[36]
Grok Imagine Image 2.0 xAI  United States August 2026 1154 1315[lower-alpha 4] $60.00 Proprietary[36]
Reve 2.1 Reve  United States July 2026 1129 1301 $200.00[36] Proprietary[37]
Muse Image Meta Platforms  United States July 2026 1111 1277 $10.00 Proprietary[36]
Nano Banana 2 Google DeepMind  United States February 2026 1122 1261 $67.00[36] Proprietary[37]
Seedream 5.0 Pro ByteDance  China 2026 1078 1257 $90.00[36] Proprietary[37]
Qwen-Image-3.0-Pro Alibaba Cloud  China July 2026 1088 1254 $40.00[36] Qwen License[38]
MAI-Image-2.5 Microsoft AI  United States June 2026 1102 1254 $48.10[36] Proprietary[37]
Nano Banana 2 Lite Google DeepMind  United States June 2026 1091 1250 $33.60[36] Proprietary[37]
Nano Banana Pro Google DeepMind  United States November 2025 1100 1246 $134.00[36] Proprietary[37]
GPT Image 1.5 OpenAI  United States December 2025 1102 1239 $133.00[36] Proprietary[37]
Ideogram 4.0 Quality Ideogram  United States June 2026 1012 1204 $60.00[36] Proprietary[37]
Luma UNI Max Luma AI  United States 2026 1012 1188 n/a Proprietary[37]
MAI-Image-2 Microsoft AI  United States 2026 1009 1183 n/a Proprietary[37]
Grok Imagine Image xAI  United States 2026 1018 1171 n/a Proprietary[37]
Recraft V4.1 Utility Pro Recraft  United States 2026 1017 1169 $210.00[36] Proprietary[37]
Cosmos3-Super-Text2Image Nvidia  United States May 2026 983 1167 n/a Apache 2.0[39]

See also

Notes

  1. ↑ An asterisk denotes an Artificial Analysis estimated score pending independent evaluation; n/a indicates no published score in the cited snapshot.
  2. ↑ 2.0 2.1 Scores shown are for each model's highest-performing reasoning effort level on Artificial Analysis; lower-effort configurations of the same model are omitted.
  3. ↑ Arena.ai scores shown are each model's best across reported effort variants on the leaderboard.
  4. ↑ 4.0 4.1 4.2 4.3 Arena Elo ratings annotated as preliminary are based on fewer than 5,000 votes at the time of collection and are subject to greater statistical uncertainty as additional votes are recorded.

References

  1. ↑ 1.0 1.1 1.2 1.3 "LLM Leaderboard - Comparison of AI Models". Artificial Analysis. https://artificialanalysis.ai/leaderboards/models. 
  2. ↑ "Artificial Analysis Intelligence Index v4.3.2". Artificial Analysis. https://artificialanalysis.ai/evaluations/artificial-analysis-intelligence-index. 
  3. ↑ Eduardo Baptista (3 August 2026). "DeepSeek's new AI model is by far the cheapest of well-known models to run, research firm says". Reuters. https://www.reuters.com/business/retail-consumer/deepseeks-new-ai-model-is-by-far-cheapest-well-known-models-run-research-firm-2026-08-03/. 
  4. ↑ 4.0 4.1 "AI startup LMArena triples its valuation to $1.7 billion in latest fundraise". Reuters. 6 January 2026. https://www.reuters.com/technology/ai-startup-lmarena-triples-its-valuation-17-billion-latest-fundraise-2026-01-06/. 
  5. ↑ Rachel Metz (17 April 2025). "Popular Chatbot-Ranking Website Is Becoming a Real Company". Bloomberg. https://www.bloomberg.com/news/articles/2025-04-17/popular-ai-ranking-website-chatbot-arena-is-becoming-a-real-company. 
  6. ↑ Rachel Metz and Katie Roof (21 May 2025). "Chatbot Arena Group Goes From Academic Project to $600 Million Startup". Bloomberg. https://www.bloomberg.com/news/articles/2025-05-21/lmarena-goes-from-academic-project-to-600-million-startup. 
  7. ↑ "Epoch Capabilities Index (ECI)". Epoch AI. https://epoch.ai/eci. 
  8. ↑ "GPT-6 Astra leads Epoch Capabilities Index with ECI of 166". Crypto Briefing. 16 September 2026. https://cryptobriefing.com/gpt-6-astra-epoch-capabilities-index/. 
  9. ↑ Kyle Wiggers (19 January 2025). "AI benchmarking organization criticized for waiting to disclose funding from OpenAI". TechCrunch. https://techcrunch.com/2025/01/19/ai-benchmarking-organization-criticized-for-waiting-to-disclose-funding-from-openai/. 
  10. ↑ "Vals Index". Vals AI. 10 September 2026. https://vals.ai/benchmarks/vals_index. 
  11. ↑ "LMArena Leaderboard". LMArena. https://lmarena.ai/leaderboard. 
  12. ↑ "Epoch Capabilities Index leaderboard: 238 AI models ranked". BenchLeader. 21 September 2026. https://www.benchleader.com/benchmarks/epoch_eci. 
  13. ↑ "Vals Index leaderboard: 55 AI models ranked". BenchLeader. 19 September 2026. https://www.benchleader.com/benchmarks/vals_index. 
  14. ↑ 14.0 14.1 14.2 14.3 14.4 14.5 "Claude - Terms and Policies". Anthropic. https://www.anthropic.com/policies/terms. 
  15. ↑ 15.0 15.1 15.2 15.3 15.4 15.5 15.6 "Terms of Use". OpenAI. https://openai.com/policies/terms-of-use/. 
  16. ↑ "Announcing the Artificial Analysis Intelligence Index v4.3". Artificial Analysis. 7 September 2026. https://artificialanalysis.ai/articles/artificial-analysis-intelligence-index-v4-3. 
  17. ↑ "Claude Fable 5 (with fallback) - Intelligence, Performance and Price Analysis". Artificial Analysis. https://artificialanalysis.ai/models/claude-fable-5. 
  18. ↑ Harshita Mary Varghese and Katie Paul (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/. ; "Introducing Muse Spark 1.1". Meta AI. 9 July 2026. https://ai.meta.com/blog/introducing-muse-spark-meta-model-api/. ; Lucas Ropek (9 July 2026). "Meta enters the crowded AI coding battle with Muse Spark 1.1". TechCrunch. https://techcrunch.com/2026/07/09/meta-enters-the-crowded-ai-coding-battle-with-muse-spark-1-1/. 
  19. ↑ 19.0 19.1 19.2 "Grok Terms". xAI. https://x.ai/legal/terms-of-service. 
  20. ↑ "MiMo-V2.6-Pro - Intelligence, Performance and Price Analysis". Artificial Analysis. https://artificialanalysis.ai/models/mimo-v2-6-pro. 
  21. ↑ 21.0 21.1 21.2 "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. ; "Alibaba plans to charge big users of its next open-source AI model, sources say". Reuters. 7 August 2026. https://www.reuters.com/business/retail-consumer/alibaba-plans-charge-big-users-its-next-open-source-ai-model-sources-say-2026-08-07/. 
  22. ↑ 22.0 22.1 22.2 "Zhipu AI Models - Hugging Face". Zhipu AI. https://huggingface.co/zai-org. 
  23. ↑ "StepFun Models". StepFun. https://www.stepfun.com/. 
  24. ↑ "Kimi K3 License". Moonshot AI. https://github.com/MoonshotAI/Kimi-K3/blob/HEAD/LICENSE. ; "Kimi K3 License". Moonshot AI. 27 July 2026. https://huggingface.co/moonshotai/Kimi-K3/blob/main/LICENSE. ; Carl Franzen (27 July 2026). "Kimi K3's full weights are here, but they're 'open' with a caveat". VentureBeat. https://venturebeat.com/technology/kimi-k3s-full-weights-are-here-but-theyre-open-with-a-caveat-what-enterprises-should-know. ; "Moonshot AI releases Kimi K3 model weights under custom licence". The Indian Express. 28 July 2026. https://indianexpress.com/article/technology/artificial-intelligence/china-moonshot-ai-kimi-k3-model-weights-custom-licence-10807478/. 
  25. ↑ 25.0 25.1 25.2 25.3 "Gemini Terms". Google DeepMind. https://policies.google.com/terms/generative-ai/use-policy. 
  26. ↑ 26.0 26.1 26.2 "DeepSeek Models - Hugging Face". DeepSeek. https://huggingface.co/deepseek-ai. 
  27. ↑ 27.0 27.1 "Sapiens AI - Singapore". Sapiens AI. https://sapiens.ai/. 
  28. ↑ "Qwen3.8-27B LICENSE - Apache License 2.0". Alibaba Cloud. https://huggingface.co/Qwen/Qwen3.8-27B/blob/main/LICENSE. 
  29. ↑ "Motif-Technologies/Motif-3 - License: MIT, released under the MIT License". Hugging Face. https://huggingface.co/Motif-Technologies/Motif-3. 
  30. ↑ "Institute of Foundation Models". MBZUAI. https://www.mbzuai.ac.ae/. 
  31. ↑ "Apodex 1.1 Moves AI Beyond Deep Research to Verifiable Execution". Redwood City, Calif.: PR Newswire. 1 September 2026. https://www.prnewswire.com/news-releases/apodex-1-1-moves-ai-beyond-deep-research-to-verifiable-execution-302866271.html. 
  32. ↑ "Apodex". Apodex. https://www.apodex.com/blog/apodex. "Apodex is developed in the United States and Singapore" 
  33. ↑ "Apodex 1.1: AA Index 44, agentic edge, and a catch". Orcarouter. 31 August 2026. https://www.orcarouter.ai/blog/apodex-1-1-explained. "Apodex, the AI company founded by Chen Tianqiao" 
  34. ↑ "Apodex 1.1: AA Index 44, agentic edge, and a catch". Orcarouter. 31 August 2026. https://www.orcarouter.ai/blog/apodex-1-1-explained. "The flagship is proprietary; the Mini 35B sibling is Apache-2.0" 
  35. ↑ Carl Franzen (21 May 2026). "Alibaba's proprietary Qwen3.7-Max can run for 35 hours autonomously". VentureBeat. https://venturebeat.com/technology/alibabas-proprietary-qwen3-7-max-can-run-for-35-hours-autonomously-and-supports-external-harnesses-like-anthropics-claude-code. 
  36. ↑ 36.00 36.01 36.02 36.03 36.04 36.05 36.06 36.07 36.08 36.09 36.10 36.11 36.12 36.13 36.14 36.15 36.16 36.17 "Text to Image Leaderboard - Top AI Image Models". Artificial Analysis. https://artificialanalysis.ai/image/leaderboard/text-to-image. 
  37. ↑ 37.00 37.01 37.02 37.03 37.04 37.05 37.06 37.07 37.08 37.09 37.10 37.11 37.12 "Text-to-Image Arena Overall". Arena.ai. 7 September 2026. https://arena.ai/leaderboard/text-to-image/overall. 
  38. ↑ "Qwen-Image Models - Hugging Face". Alibaba Cloud. https://huggingface.co/Qwen. 
  39. ↑ "Cosmos Models - Hugging Face". Nvidia. https://huggingface.co/nvidia. 



Retrieved from "https://handwiki.org/wiki/index.php?title=Comparison_of_generative_AI_models&oldid=4633707"