AI model leaderboard
Text, image, and video models ranked on one page — an independent quality score on one axis, the price OttoPort actually bills on the other. Every model in the tables is one API key away.
OttoPort does not run model evaluations. Quality and speed are copied from published third-party leaderboards and cited per row, with the date each was read — 55 of the 126 models carry one, and the rest show "—" rather than an estimate. Prices are read from the same catalog the gateway meters requests against, so the table cannot quote a rate you are not billed.
- Top text model
- Claude Opus 5.5
- 58 Intelligence Index · $8.86 / 1M tokens
- Top image model
- GPT Image 2.5 Flare
- 1189 Arena Elo · $0.026 / image
- Top video model
- Seedance 2.5
- 1482 Arena Elo · $0.35 / second
Ranked by modality
Scores only mean something within a modality — an intelligence index and an Elo rating measure different things by different methods — so each table ranks only against itself. Sort any column.
Language models, ranked by the Artificial Analysis Intelligence Index, against a blended price at three input tokens per output token.
| # | Model | ||||
|---|---|---|---|---|---|
| 1 | Claude Opus 5.51Anthropic | 58 | $8.86 / 1M tokens | 92 t/s | 6.5 |
| 2 | GPT-6 AstraOpenAI | 53 | $22.16 / 1M tokens | 53 t/s | 2.4 |
| 3 | Claude Fable 5.1Anthropic | 53 | $22.16 / 1M tokens | 65 t/s | 2.4 |
| 4 | Claude Opus 5Anthropic | 51 | $11.08 / 1M tokens | 49 t/s | 4.6 |
| 5 | Claude Fable 5Anthropic | 50 | $22.16 / 1M tokens | 65 t/s | 2.3 |
| 6 | GPT-6 Sol2OpenAI | 48 | $4.44 / 1M tokens | 93 t/s | 10.8 |
| 7 | GPT-5.6 SolOpenAI | 47 | $4.44 / 1M tokens | 65 t/s | 10.6 |
| 8 | Grok 4.73xAI | 46 | $3.33 / 1M tokens | 45 t/s | 13.8 |
| 9 | GLM 5.3Z.ai | 45 | $1.87 / 1M tokens | 67 t/s | 24.1 |
| 10 | Grok 4.6xAI | 44 | $3.33 / 1M tokens | 58 t/s | 13.2 |
| 11 | Kimi K3Moonshot AI | 44 | $6.64 / 1M tokens | 35 t/s | 6.6 |
| 12 | GLM 5.3 FlashZ.ai | 42 | $0.263 / 1M tokens | 43 t/s | 160 |
| 13 | GPT-5.6 TerraOpenAI | 42 | $4.99 / 1M tokens | 99 t/s | 8.4 |
| 14 | Gemini 3.8 FlashGoogle | 41 | $1.66 / 1M tokens | 332 t/s | 24.7 |
| 15 | Qwen3.8 MaxQwen | 40 | $3.33 / 1M tokens | 40 t/s | 12.0 |
| 16 | DeepSeek V4.1 Flash4DeepSeek | 39 | $0.2407 / 1M tokens | 233 t/s | 162 |
| 17 | Gemini 3.7 FlashGoogle | 39 | $1.66 / 1M tokens | 289 t/s | 23.5 |
| 18 | Grok 4.5xAI | 39 | $3.33 / 1M tokens | 60 t/s | 11.7 |
| 19 | GPT-5.6 LunaOpenAI | 38 | $0.499 / 1M tokens | 115 t/s | 76.2 |
| 20 | Claude Sonnet 5Anthropic | 38 | $4.44 / 1M tokens | 80 t/s | 8.6 |
| 21 | GPT-6 Luna5OpenAI | 37 | $0.2218 / 1M tokens | 145 t/s | 167 |
| 22 | GLM 5.2Z.ai | 34 | $1.19 / 1M tokens | 71 t/s | 28.5 |
| 23 | DeepSeek V4 ProDeepSeek | 31 | $1.33 / 1M tokens | 86 t/s | 23.4 |
| 24 | Gemini 3.1 Pro Preview6Google | 30 | $4.99 / 1M tokens | 108 t/s | 6.0 |
| 25 | Kimi K2.7 CodeMoonshot AI | 26 | $1.49 / 1M tokens | 64 t/s | 17.5 |
| 26 | Gemini 3.5 Flash-LiteGoogle | 22 | $0.9415 / 1M tokens | 350 t/s | 23.4 |
| 27 | Claude Haiku 4.5Anthropic | 18 | $2.22 / 1M tokens | 85 t/s | 8.1 |
| 28 | DeepSeek V4 FlashDeepSeek | — | $0.1263 / 1M tokens | — | — |
| 29 | GPT-6 Luna ProOpenAI | — | $0.2218 / 1M tokens | — | — |
| 30 | Claude Haiku 5.5Anthropic | — | $0.2218 / 1M tokens | — | — |
| 31 | Qwen3.8 FlashQwen | — | $0.2548 / 1M tokens | — | — |
| 32 | GPT-5.4 nanoOpenAI | — | $0.5115 / 1M tokens | — | — |
| 33 | Gemini 3.1 Flash-LiteGoogle | — | $0.6228 / 1M tokens | — | — |
| 34 | Kimi K2.5Moonshot AI | — | $0.828 / 1M tokens | — | — |
| 35 | Gemini 2.5 FlashGoogle | — | $0.9415 / 1M tokens | — | — |
| 36 | Kimi K2 0711Moonshot AI | — | $1.05 / 1M tokens | — | — |
| 37 | Kimi K2 ThinkingMoonshot AI | — | $1.13 / 1M tokens | — | — |
| 38 | Kimi K2 0905Moonshot AI | — | $1.13 / 1M tokens | — | — |
| 39 | Kimi K2.6Moonshot AI | — | $1.41 / 1M tokens | — | — |
| 40 | GPT-5.4 miniOpenAI | — | $1.87 / 1M tokens | — | — |
| 41 | DeepSeek V4 Pro 0813DeepSeek | — | $1.87 / 1M tokens | — | — |
| 42 | Gemini 3.5 FlashGoogle | — | $3.74 / 1M tokens | — | — |
| 43 | GPT-6.1 SolOpenAI | — | $4.44 / 1M tokens | — | — |
| 44 | GPT-6.1 Sol ProOpenAI | — | $4.44 / 1M tokens | — | — |
| 45 | GPT-6 Sol ProOpenAI | — | $4.44 / 1M tokens | — | — |
| 46 | Claude Sonnet 5.5Anthropic | — | $4.44 / 1M tokens | — | — |
| 47 | GPT-4oOpenAI | — | $4.85 / 1M tokens | — | — |
| 48 | Video UnderstandingOttoPort | — | $4.99 / 1M tokens | — | — |
| 49 | GPT-5.2OpenAI | — | $5.33 / 1M tokens | — | — |
| 50 | GPT-5.4OpenAI | — | $6.23 / 1M tokens | — | — |
| 51 | Claude Opus 4.8Anthropic | — | $11.08 / 1M tokens | — | — |
| 52 | GPT-5.5OpenAI | — | $12.46 / 1M tokens | — | — |
| 53 | Sonar Deep ResearchPerplexity | — | $13.30 / 1M tokens | — | — |
| 54 | GPT-6 Astra ProOpenAI | — | $22.16 / 1M tokens | — | — |
- 1Claude Opus 5.5 Index is the max-effort run; AA publishes no speed for it, so speed is the xhigh run's.
- 2GPT-6 Sol Scored at max reasoning effort.
- 3Grok 4.7 Scored at xhigh reasoning effort.
- 4DeepSeek V4.1 Flash Scored at max reasoning effort.
- 5GPT-6 Luna Scored at max reasoning effort.
- 6Gemini 3.1 Pro Preview Scored as Gemini 3.1 Pro; the catalog entry is the preview build of the same model.
Image models, ranked by Elo from blind head-to-head votes in the Artificial Analysis text-to-image arena, against the price of one image.
| # | Model | |||
|---|---|---|---|---|
| 1 | GPT Image 2.5 Flare1OpenAI | 1189 | $0.026 / image | 7,269 |
| 2 | GPT Image 2.52OpenAI | 1183 | $0.026 / image | 7,038 |
| 3 | GPT Image 23OpenAI | 1173 | $0.07 / image | 2,471 |
| 4 | Nano Banana 2Google | 1122 | $0.088 / image | 1,386 |
| 5 | GPT Image 1.54OpenAI | 1104 | $0.044 / image | 2,364 |
| 6 | Nano Banana ProGoogle | 1099 | $0.165 / image | 600 |
| 7 | Nano Banana 2 LiteGoogle | 1088 | $0.044 / image | 2,000 |
| 8 | Qwen Image 3.0 ProQwen | 1087 | $0.044 / image | 1,977 |
| 9 | Seedream 5.05ByteDance | 1082 | $0.05 / image | 1,640 |
| 10 | Qwen Image 3.0Qwen | 1073 | $0.033 / image | 2,212 |
| 11 | Seedream 5.0 LiteByteDance | 1000 | $0.039 / image | 0.00 |
| 12 | Wan 2.7 Pro ImageAlibaba | 979 | $0.083 / image | 0.00 |
| 13 | GPT Image 1 mini6OpenAI | 919 | $0.022 / image | 0.00 |
| 14 | Midjourney V77Midjourney | 897 | $0.088 / image | 0.00 |
| 15 | Qwen Image LayeredQwen | — | $0.028 / layer | — |
| 16 | Seedream 4.7ByteDance | — | $0.033 / image | — |
| 17 | Seedream 4.7 SequentialByteDance | — | $0.033 / image | — |
| 18 | FLUX.2 ProBlack Forest Labs | — | $0.033 / image | — |
| 19 | Seedream 5.0 LayersByteDance | — | $0.037 / layer | — |
| 20 | Qwen Image AnglesQwen | — | $0.04 / image | — |
| 21 | Ideogram 4.5Ideogram | — | $0.066 / image | — |
| 22 | Nano Banana 2.1Google | — | $0.088 / image | — |
| 23 | Midjourney V8Midjourney | — | $0.088 / image | — |
| 24 | Midjourney V8.2Midjourney | — | $0.088 / image | — |
| 25 | Midjourney V8.1Midjourney | — | $0.095 / image | — |
- 1GPT Image 2.5 Flare Arena entry: GPT Image 2.5 Flare (max quality setting).
- 2GPT Image 2.5 Arena entry: GPT Image 2.5 Sunburst (max quality setting).
- 3GPT Image 2 Arena entry: GPT Image 2 at the high quality setting.
- 4GPT Image 1.5 Arena entry: GPT Image 1.5 at the high quality setting.
- 5Seedream 5.0 Arena entry: Seedream 5.0 Pro.
- 6GPT Image 1 mini Arena entry: GPT Image 1 Mini at the medium quality setting.
- 7Midjourney V7 Arena entry: Midjourney v7 Alpha.
Video models, ranked by Elo from blind head-to-head votes in the arena.ai text-to-video arena, against the price of one second of finished video.
| # | Model | |||
|---|---|---|---|---|
| 1 | Seedance 2.51ByteDanceOttoPort pick | 1482 | $0.35 / second | 1,377 |
| 2 | Gemini Omni 1.1 FlashGoogle | 1515 | $0.138 / second | 3,732 |
| 3 | Gemini Omni FlashGoogle | 1511 | $0.156 / second | 3,276 |
| 4 | Wan 3.0 Text to Video2Alibaba | 1494 | $0.20 / second | 2,470 |
| 5 | FLUX 33Black Forest Labs | 1494 | $0.213 / second | 2,319 |
| 6 | Seedance 2.04ByteDance | 1479 | $0.227 / second | 2,110 |
| 7 | MiniMax H3MiniMax | 1462 | $0.163 / second | 2,834 |
| 8 | Veo 3.15Google | 1364 | $0.528 / second | 689 |
| 9 | Veo 3.1 Fast6Google | 1362 | $0.132 / second | 2,742 |
| 10 | Wan 2.7 Text to VideoAlibaba | 1341 | $0.125 / second | 2,728 |
| 11 | Wan 2.6 Text to VideoAlibaba | 1328 | $0.088 / second | 3,727 |
| 12 | Wan 2.5 Text to Video7Alibaba | 1246 | $0.063 / second | 3,905 |
| 13 | MiniMax Hailuo 2.3MiniMax | 1205 | $0.07 / second | 2,929 |
| 14 | MiniMax Hailuo 028MiniMax | 1181 | $0.07 / second | 2,586 |
| 15 | Kling LipsyncKuaishou | — | $0.018 / second | — |
| 16 | Wan 2.6 Image to Video FlashAlibaba | — | $0.031 / second | — |
| 17 | Seedance 2.0 MiniByteDance | — | $0.043 / second | — |
| 18 | MiniMax Hailuo 2.3 FastMiniMax | — | $0.048 / second | — |
| 19 | Seedance 1.5 Pro FastByteDance | — | $0.05 / second | — |
| 20 | Seedance 1.5 ProByteDance | — | $0.063 / second | — |
| 21 | Wan 2.5 Image to VideoAlibaba | — | $0.063 / second | — |
| 22 | Video Depth AnythingByteDance | — | $0.063 / second | — |
| 23 | Sync Lipsync 2Sync Labs | — | $0.063 / second | — |
| 24 | Veo 3.1 LiteGoogle | — | $0.066 / second | — |
| 25 | Kling Avatar 2.0Kuaishou | — | $0.07 / second | — |
| 26 | Wan 2.6 Image to VideoAlibaba | — | $0.088 / second | — |
| 27 | Kling 2.6 Motion ControlKuaishou | — | $0.088 / second | — |
| 28 | Sync Lipsync 2 ProSync Labs | — | $0.104 / second | — |
| 29 | Kling 3.0Kuaishou | — | $0.105 / second | — |
| 30 | Kling O3Kuaishou | — | $0.105 / second | — |
| 31 | Seedance 2.0 FastByteDance | — | $0.11 / second | — |
| 32 | Kling 3.0 TurboKuaishou | — | $0.119 / second | — |
| 33 | Kling O1Kuaishou | — | $0.119 / second | — |
| 34 | Wan 2.7 Image to VideoAlibaba | — | $0.125 / second | — |
| 35 | Kling O3 ProKuaishou | — | $0.14 / second | — |
| 36 | Kling 2.6 Pro Motion ControlKuaishou | — | $0.14 / second | — |
| 37 | Kling Avatar 2.0 ProKuaishou | — | $0.144 / second | — |
| 38 | Kling 3.0 Motion ControlKuaishou | — | $0.158 / second | — |
| 39 | MiniMax H3 MaxMiniMax | — | $0.178 / second | — |
| 40 | Kling 3.0 ProKuaishou | — | $0.179 / second | — |
| 41 | Wan 3.0 Image to VideoAlibaba | — | $0.20 / second | — |
| 42 | OmniHuman 1.5ByteDance | — | $0.20 / second | — |
| 43 | Kling 3.0 Pro Motion ControlKuaishou | — | $0.21 / second | — |
| 44 | Wan 3.0 Prime Text to VideoAlibaba | — | $0.315 / second | — |
| 45 | Wan 3.0 Prime Image to VideoAlibaba | — | $0.315 / second | — |
| 46 | Kling 3.0 4KKuaishou | — | $0.446 / second | — |
| 47 | Kling O3 4KKuaishou | — | $0.525 / second | — |
Seedance 2.5 is placed first by OttoPort, not by a leaderboard. OttoPort's pick, not a benchmark result: ByteDance's current flagship — native 30-second clips, 4K, up to 50 reference images. Its measured Elo stays in the column unchanged; only the position is OttoPort's call. arena.ai has it 6th at 1482 and Artificial Analysis has not scored it at all, because both boards are still accumulating votes on a model that shipped after they last settled.
- 1Seedance 2.5 Arena entry: dreamina-seedance-2.5-720p. Not yet ranked on the Artificial Analysis board.
- 2Wan 3.0 Text to Video Arena entry: wan3.0. The Prime tier is not separately ranked.
- 3FLUX 3 Arena entry: flux-3-video.
- 4Seedance 2.0 Arena entry: dreamina-seedance-2.0-720p.
- 5Veo 3.1 Arena entry: veo-3.1-audio at 720p; the 1080p entry scores 1363.
- 6Veo 3.1 Fast Arena entry: veo-3.1-fast-audio at 720p.
- 7Wan 2.5 Text to Video Arena entry: wan2.5-t2v-preview, the same upstream build this id routes to.
- 8MiniMax Hailuo 02 Arena entry: hailuo-02-standard; the Pro tier scores 1198.
Other boards worth reading
One ranking cannot cover every job a model is hired for. These boards measure a different task on their own scale — they are reproduced here, not folded into the tables above.
Video editing
arena.ai Video Edit Arena (Elo) · read 2026-09-16Editing a clip the model is handed — restyling it, replacing an object, extending it — rather than generating one from a prompt. A different task, a separate Elo pool, and a visibly different order.
- 1Wan 3.01414
- 2Dreamina Seedance 2.5 720p1410
- 3MiniMax H31392
- 4Gemini Omni Flash1367
- 5Dreamina Seedance 2.0 720p1365
- 6HappyHorse 1.0not on OttoPort1307
- 7grok-imagine-videonot on OttoPort1258
- 8Kling O3 Pronot on OttoPort1255
- 9Kling O1 Pronot on OttoPort1197
- 10Runway Gen-4 Alephnot on OttoPort1182
How these rankings are built
Quality and speed: not ours
Language models carry the Artificial Analysis Intelligence Index; image and video models carry Arena Elo from blind head-to-head votes — image from Artificial Analysis, video from arena.ai, whose board covers the newest models in the catalog. Elo pools are not interchangeable, so each column comes from exactly one board. Every row cites the one it came from and the date it was read. Sources:
Price: the rate you are billed
Text models are priced at a blended 3:1 mix of input to output tokens — the ratio the index's own price axis uses. Image models are priced per image at their base resolution, video models per second of finished output. Every figure comes from the live catalog, not a copy of it.
Value: quality per dollar
The quality score divided by the comparable price. For Elo-scored image and video models it counts points above the arena's 1000-point baseline, because Elo has no meaningful zero — a model at 1100 is not "10% better" than one at 1000, but it is measurably ahead of it.
Picks: the one opinion here
Arenas lag the market — a model can ship, lead its category, and wait months to be scored. Where that happens OttoPort pins the model to the top of the default view, badged as a pick with the reason printed under the table. A pick never writes a score into a cell, and it disappears the moment you sort by any column. OttoPort sells these models; read the picks knowing that, and the measured columns knowing OttoPort did not produce them.
Prices for every model, including speech and music, are on the pricing comparison, and the full catalog is at /models.
Common questions
What is the best AI model right now?
For text, the highest-scoring model on OttoPort is Claude Opus 5.5 at 58 on the Artificial Analysis Intelligence Index, at $8.86 / 1M tokens. For images it is GPT Image 2.5 Flare at 1189 Elo in the text-to-image arena, at $0.026 / image. For video, OttoPort places Seedance 2.5 first at $0.35 / second — that position is OttoPort's own pick rather than a benchmark result, and the arena that has scored it puts it at 1482 Elo. The highest arena-scored video model on OttoPort is Seedance 2.5 at 1482 Elo. Scores are third-party measurements, cited per row; prices are OttoPort's own published rates.
Does OttoPort put its own models first?
OttoPort resells every model on this page, so the ranking is deliberately built from figures OttoPort does not control: third-party benchmark scores, the published rate card, and metered request counts. The one exception is an editorial pick — a model OttoPort places above its measured position because the boards have not caught up with it yet. A pick never invents a score, it is badged "OttoPort pick" on the row with the reason printed under the table, and it holds only in the default view: sort by price, speed, value, or usage and it drops to wherever its numbers put it.
Where do these scores come from?
OttoPort does not run model evaluations, so every quality score on this page is copied from a published third-party leaderboard and cited on the row it appears in, with the date it was read. 55 of the 126 ranked models carry a score; the rest are models the sources had not yet published a figure for, and they render as "—" rather than being estimated.
Which AI model gives the best value for money?
Sort any of the three tables by "Value" — quality per dollar at the model's published rate. For Elo-scored image and video models, value counts Elo points above the arena's 1000-point baseline, because Elo has no meaningful zero. On text, GPT-6 Luna currently leads on that measure at $0.2218 / 1M tokens.
Are text, image, and video scores comparable to each other?
No. Text models carry an intelligence index out of roughly 100; image and video models carry Arena Elo, which runs in the 450-1350 range and comes from blind head-to-head human votes. The two scales measure different things by different methods, so each table ranks only within itself.