AI model and media leaderboards from Artificial Analysis
Independent third-party measurements of language models, image, video and speech generators. Snapshot of 2026-10-27. Not our tests.
Language models by intelligence index
The 20 highest-scoring models of 60 tracked here (best setting of each). A composite of several benchmarks, with blended price and output speed.
Language model leaderboard
| # | Model | Maker | Intelligence index | Blended price per million tokens | Output tokens per second |
|---|---|---|---|---|---|
| 1 | Claude Opus 5.5 | Anthropic | 57.6 | $8 | 97 |
| 2 | Claude Sonnet 5.5 | Anthropic | 56 | $4 | 141 |
| 3 | Muse Wonder Full | Meta | 55.8 | n/a | n/a |
| 4 | Kimi K3.1 Pro | Kimi | 55.1 | $4 | 38 |
| 5 | Claude Fable 5.1 | Anthropic | 53.4 | $20 | 66 |
| 6 | GPT-6 Astra | OpenAI | 52.7 | $20 | 68 |
| 7 | Gemini 4 Argon | 52.6 | $4 | 82 | |
| 8 | Muse Wonder | Meta | 52.4 | $8 | 173 |
| 9 | GPT-6.1 Sol | OpenAI | 51.8 | $4 | 75 |
| 10 | Claude Opus 5 | Anthropic | 50.8 | $10 | n/a |
| 11 | Claude Fable 5 | Anthropic | 49.6 | $20 | n/a |
| 12 | GPT-6.1 Terra | OpenAI | 48.1 | $2 | 91 |
| 13 | Muse Spark 1.3 | Meta | 48.1 | $2 | n/a |
| 14 | GPT-6 Sol | OpenAI | 47.6 | $4 | n/a |
| 15 | GPT-5.6 Sol | OpenAI | 47 | $8 | n/a |
| 16 | Grok 4.7 | SpaceXAI | 46.4 | $3 | 79 |
| 17 | MiMo-V2.6-Pro | Xiaomi | 46.3 | $0.544 | 50 |
| 18 | Qwen3.8 Max | Alibaba | 45.4 | $3 | 40 |
| 19 | DeepSeek V4.1 Flash 1017 | DeepSeek | 45.1 | $0.525 | 162 |
| 20 | GLM-5.3 | Z AI | 44.8 | $2.15 | 82 |
Models we have a review for link to it. Blended price assumes 3 parts input to 1 part output. "n/a" means Artificial Analysis lists no value.
Text-to-speech
Which voices listeners prefer in blind comparisons.
Text-to-speech
| Rank | Model | Maker | Elo | 95% interval |
|---|---|---|---|---|
| 1 | Eleven v4 | ElevenLabs | 1336 | -15/15 |
| 2 | Eleven v4 Turbo | ElevenLabs | 1333 | -16/16 |
| 3 | Qwen-Audio-3.1-TTS-Plus | Alibaba | 1306 | -15/15 |
| 4 | Sonic 3.6 | Cartesia | 1294 | -14/14 |
| 5 | Gemini 3.8 Flash TTS | 1283 | -13/13 | |
| 6 | Qwen-Audio-3.0-TTS-Plus | Alibaba | 1269 | -13/13 |
| 7 | Realtime TTS-2 | Inworld | 1267 | -14/14 |
| 8 | Gemini 3.8 Flash-Lite TTS | 1259 | -13/13 | |
| 9 | Simba 3.2 | SpeechifyAI | 1249 | -12/12 |
| 10 | Fish Audio Drama 3 Preview | Fish Audio | 1245 | -14/14 |
| 15 | v3 Conversational | ElevenLabs | 1214 | -12/12 |
| 20 | Eleven v3 | ElevenLabs | 1187 | -10/10 |
| 27 | Falcon 2 | Murf AI | 1157 | -12/12 |
| 43 | Turbo v2.5 | ElevenLabs | 1095 | -10/10 |
| 44 | Multilingual v2 | ElevenLabs | 1093 | -10/10 |
| 50 | Flash v2.5 | ElevenLabs | 1077 | -10/10 |
| 65 | Flux TTS | Deepgram | 1050 | -13/13 |
| 82 | Murf Speech Gen 2 | Murf AI | 986 | -12/12 |
Top 10 of 98 models, plus every ElevenLabs, Murf AI or Deepgram model ranked lower. Ranked by Elo from blind votes.
Text-to-image
Image generators ranked by blind votes.
Text-to-image
| Rank | Model | Maker | Elo | 95% interval |
|---|---|---|---|---|
| 1 | GPT Image 2.5 Sunburst (max) | OpenAI | 1222 | -10/10 |
| 2 | GPT Image 2.5 Flare (max) | OpenAI | 1207 | -10/10 |
| 3 | GPT Image 2 (high) | OpenAI | 1186 | -11/11 |
| 4 | Nano Banana 2.1 | 1176 | -12/12 | |
| 5 | MAI-Image-2.6 | Microsoft AI | 1166 | -12/12 |
| 6 | Grok Imagine Image 2.0 | SpaceXAI | 1166 | -12/12 |
| 7 | Nano Banana 2 (Gemini 3.1 Flash Image) | 1140 | -10/10 | |
| 8 | Muse Image | Meta | 1127 | -11/11 |
| 9 | GPT Image 1.5 (high) | OpenAI | 1123 | -11/11 |
| 10 | MAI-Image-2.6-Flash | Microsoft AI | 1119 | -12/12 |
Top 10 of 106 models. Ranked by Elo from blind votes.
Image editing
How well models follow editing instructions, by blind votes.
Image editing
| Rank | Model | Maker | Elo | 95% interval |
|---|---|---|---|---|
| 1 | GPT Image 2.5 Sunburst (max) | OpenAI | 1199 | -11/11 |
| 2 | MAI-Image-2.7-Preview | Microsoft AI | 1193 | -15/15 |
| 3 | GPT Image 2.5 Flare (max) | OpenAI | 1182 | -11/11 |
| 4 | MAI-Image-2.6 | Microsoft AI | 1151 | -12/12 |
| 5 | Krea 3 (Ultra) | Krea | 1151 | -14/14 |
| 6 | Nano Banana 2.1 | 1151 | -13/13 | |
| 7 | MAI-Image-2.6-Flash | Microsoft AI | 1144 | -13/13 |
| 8 | GPT Image 2 (high) | OpenAI | 1136 | -13/13 |
| 9 | Seedream 5.0 Pro | ByteDance Seed | 1131 | -13/13 |
| 10 | Muse Image | Meta | 1129 | -12/12 |
Top 10 of 77 models. Ranked by Elo from blind votes.
Text-to-video
Video generators ranked by blind votes from text prompts.
Text-to-video
| Rank | Model | Maker | Elo | 95% interval |
|---|---|---|---|---|
| 1 | HeyGen Video 1 | HeyGen | 1143 | -16/16 |
| 2 | Wan 3.0 | Alibaba | 1133 | -17/17 |
| 3 | MiniMax H3 (768p) | MiniMax | 1120 | -17/17 |
| 4 | MiniMax H3 Max | Fal | 1118 | -17/17 |
| 5 | Dreamina Seedance 2.5 | ByteDance Seed | 1104 | -17/17 |
| 6 | FLUX 3 | Black Forest Labs | 1091 | -17/17 |
| 7 | Dreamina Seedance 2.0 | ByteDance Seed | 1082 | -17/17 |
| 8 | Gemini Omni Flash 1.1 | 1067 | -17/17 | |
| 9 | Agnes-Video-2.5 | Sapiens AI | 1054 | -17/17 |
| 10 | Grok Imagine Video 1.5 | SpaceXAI | 1034 | -17/17 |
| 24 | Runway Gen-4.5 (720p) | Runway | 909 | -18/18 |
Top 10 of 32 models, plus every Runway model ranked lower. Ranked by Elo from blind votes.
Image-to-video
Video generators ranked by blind votes when animating a still image.
Image-to-video
| Rank | Model | Maker | Elo | 95% interval |
|---|---|---|---|---|
| 1 | Gemini Omni Flash | 1368 | -11/11 | |
| 2 | Bach 1.0 Pro | Video Rebirth | 1362 | -11/11 |
| 3 | Wan 3.0 | Alibaba | 1362 | -13/13 |
| 4 | MiniMax H3 | MiniMax | 1356 | -11/11 |
| 5 | Vimoo 1.0 | Vimoo | 1348 | -14/14 |
| 6 | Dreamina Seedance 2.0 720p | ByteDance Seed | 1339 | -9/9 |
| 7 | PixVerse V6 | PixVerse | 1339 | -10/10 |
| 8 | Grok Imagine Video 1.5 | SpaceXAI | 1331 | -11/11 |
| 9 | grok-imagine-video | SpaceXAI | 1328 | -9/9 |
| 10 | HappyHorse-1.1 | Alibaba-ATH | 1314 | -10/10 |
| 23 | Runway Gen-4.5 (720p) | Runway | 1265 | -10/10 |
| 53 | Midjourney V1 | Midjourney | 1165 | -11/11 |
| 64 | Runway Gen 4 | Runway | 1088 | -11/11 |
| 69 | Runway Gen 3 Alpha | Runway | 1017 | -12/12 |
| 72 | Runway Gen 3 Alpha Turbo | Runway | 1003 | -12/12 |
Top 10 of 75 models, plus every Runway or Midjourney model ranked lower. Ranked by Elo from blind votes.
Data and methodology: Artificial Analysis free API, retrieved 2026-10-27. Used with attribution as its terms require. See also the frontier gap report: the Intelligence Index v5.0 compared by lab, country and API access.
Frequently asked questions
Are these your test results?
No. Artificial Analysis ran every measurement. We read its published data through its API and show it with attribution. Our own editor's ratings are separate and are never mixed into these tables.
What is an Elo score here?
A rating built from blind head-to-head votes: listeners pick the better output without knowing the model. A higher Elo means it won more votes, with the 95% interval showing the uncertainty. It reflects listener preference, not accuracy for your task.
Why does a tool's own page show a different number?
Leaderboards change as votes arrive and models are added. Each review records the figure and the date it was read. This page carries the snapshot date, so compare dates before comparing numbers.
How often is this page updated?
It is a dated snapshot, refreshed when we re-pull the data. The date at the top of the page says when it was last read.