AI model comparison
Chat, image and video AI models
Chat, image and video AI models
in one place.
Language models from Anthropic, OpenAI, Google, Meta, xAI, DeepSeek, Alibaba, Moonshot, Mistral and NVIDIA: release dates, pricing, context window and published benchmark results.
Data as of 17 September 2026 · AI tools guide →
109models
10companies
96.0%GPQA Diamond · GPT-6 Astra
24open-weight models
Scores over time
Each dot is a model. Click a company to show or hide it. Models without a published score are not plotted.
PhD-level science questions
Release timeline by year
20235 models
- GPT-3.5 Turbo
- GPT-4
- Claude
- Claude 2
- Claude 2.1
202424 models
- Gemini 1.0 Pro
- Grok-1.5
- Claude 3 Opus
- Claude 3 Sonnet
- Claude 3 Haiku
- GPT-4 Turbo
- GPT-4o
- Gemini 1.5 Pro
- Gemini 1.5 Flash
- DeepSeek-V2.5
- Codestral 22B
- Claude 3.5 Sonnet
- GPT-4o mini
- Llama 3.1 405B
- Mistral Large 2
- Grok-2
- Qwen2.5 72B
- Llama 3.1 Nemotron 70B
- Claude 3.5 Sonnet (new)
- Claude 3.5 Haiku
- o1
- Gemini 2.0 Flash
- Llama 3.3 70B
- DeepSeek-V3
202537 models
- o3-mini
- DeepSeek-R1
- Mistral Small 3
- GPT-4.5
- Grok-3
- Claude 3.7 Sonnet
- Gemma 3 27B
- GPT-4.1
- o3
- o4-mini
- Llama 4 Maverick
- Qwen3 32B
- Llama 3.1 Nemotron Ultra 253B
- Gemini 2.5 Pro
- Gemini 2.5 Flash
- Claude Opus 4
- Claude Sonnet 4
- Grok-4
- Grok-4 Heavy
- Qwen3-Coder 480B
- Kimi K2
- GPT OSS 120B
- GPT-5
- Nemotron Nano 9B v2
- Claude Opus 4.1
- Claude Sonnet 4.5
- Claude Haiku 4.5
- GPT-5.1
- Gemini 3 Pro
- Grok-4.1 Fast
- Claude Opus 4.5
- GPT-5.2
- Gemini 3 Flash
- DeepSeek-V3.2
- Qwen3 Max
- Mistral Large 3
- Nemotron 3 Nano (30B)
202643 models
- Kimi K2.5
- GPT-5.3 Codex
- Gemini 3.1 Pro
- Qwen3.5-397B-A17B
- Claude Opus 4.6
- Claude Sonnet 4.6
- GPT-5.4
- GPT-5.4 mini
- Qwen3.6 Plus
- Mistral Small 4
- Nemotron 3 Super (120B)
- GPT-5.5
- Gemma 4 31B
- Muse Spark
- DeepSeek-V4-Pro-Max
- DeepSeek-V4-Flash-Max
- Kimi K2.6
- Mistral Medium 3.5
- Claude Mythos Preview
- Claude Opus 4.7
- Gemini 3.5 Flash
- Grok 4.3
- Qwen3.7 Max
- Claude Opus 4.8
- Nemotron 3 Ultra (550B)
- Claude Fable 5
- Claude Sonnet 5
- GPT-5.6 Sol
- GPT-5.6 Terra
- GPT-5.6 Luna
- Gemini 3.6 Flash
- Muse Spark 1.1
- Grok 4.5
- Kimi K3
- Claude Opus 5
- Grok 4.6
- Qwen3.8 Max
- Nemotron 3.5 Lightning (30B)
- GPT-6 Astra
- Gemini 3.8 Flash
- Muse Spark 1.3
- DeepSeek-V4.1-Flash
- Claude Fable 5.1
Capabilities: who can do what, and how well?
Some models cannot read images, use tools or operate a computer. Among those that can, scores vary widely.
👁
Vision
Can read and understand photos, screenshots, documents and charts.
✓ 64✕ 25? 20
MMMU-Pro · Vision: multi-discipline questions with charts, diagrams and images (hard version) · models since Sep 2025
Best
- Gemini 3.5 Flash83.6
- GPT-5.583.2
- GPT-5.6 Sol83.0
Weakest
- DeepSeek-V4.1-Flash56.5
- Claude Sonnet 4.675.6
- GPT-5.4 mini76.6
Without this capability: Nemotron 3 Nano (30B), DeepSeek-V3.2, Nemotron Nano 9B v2, GPT OSS 120B, Kimi K2, Qwen3-Coder 480B, Llama 3.1 Nemotron Ultra 253B, Qwen3 32B +17
🛠
Agent / tool use
Can run tasks on its own by using tools such as web search, code execution, APIs and apps.
✓ 93✕ 4? 12
Terminal-Bench 2.x · Agentic tasks in the terminal · models since Sep 2025
Best
- Gemini 3.8 Flash90.8
- Claude Opus 589.1
- GPT-5.6 Sol88.8
Weakest
- Claude Haiku 4.541.8
- DeepSeek-V3.246.4
- Gemini 3 Flash47.6
Without this capability: DeepSeek-R1, Grok-1.5, Claude 2, Claude
🖥
Computer use
Can see the screen and operate a computer or browser with mouse and keyboard like a person.
✓ 33✕ 36? 40
OSWorld-Verified · Using a computer (screen, mouse, keyboard) · models since Sep 2025
Best
- Qwen3.8 Max86.1
- Claude Fable 585.0
- Kimi K384.8
Weakest
- Qwen3.6 Plus62.5
- GPT-5.3 Codex64.7
- GPT-5.4 mini72.1
Without this capability: Nemotron 3 Nano (30B), DeepSeek-V3.2, Nemotron Nano 9B v2, GPT OSS 120B, Kimi K2, Qwen3-Coder 480B, Llama 3.1 Nemotron Ultra 253B, Qwen3 32B +28
All models
By capability:
Click a column header to sort · 109 models shown
| Model | Company | Released | Capabilities | Humanity's Last Exam | GPQA Diamond | SWE-bench Verified | 🛠 Terminal-Bench 2.x | 🛠 τ²-Bench (Telecom) | 🖥 OSWorld-Verified | 👁 MMMU-Pro | 👁 MMMU | Price $ / MTok (in / out) | Context | Open weights |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Claude Fable 5.1 | Anthropic |
👁✓
—
🛠✓
55.8TB 4.0
🖥✓
—
|
65.0S | — | — | — | — | — | — | — | $10 / $50 | 1M | — | |
| DeepSeek-V4.1-Flash | DeepSeek |
👁✓
56.5MMMU-Pro
🛠✓
—
🖥?
—
|
36.8S | 90.9S | — | — | — | — | 56.5D | — | $0.22 / $0.66 | 1M | — | |
| Muse Spark 1.3 | Meta |
👁?
—
🛠✓
—
🖥?
—
|
— | — | — | — | — | — | — | — | $0.1 / $0.2 | 1M | — | |
| Gemini 3.8 Flash |
👁✓
—
🛠✓
90.8TB 2.x
🖥✓
59.0OSWorld 2.0
|
— | — | — | 90.8D | — | — | — | — | $0.75 / $3.75 | 1M | — | ||
| GPT-6 Astra | OpenAI |
👁✓
92.7ScreenSpot-Pro
🛠✓
57.7TB 4.0
🖥✓
72.6OSWorld 2.0
|
— | 96.0S | — | — | — | — | — | — | $10 / $50 | 1.1M | — | |
| Nemotron 3.5 Lightning (30B) | NVIDIA |
👁?
—
🛠✓
—
🖥?
—
|
— | — | — | — | — | — | — | — | $0.08 / $0.2 | 262K | Yes | |
| Qwen3.8 Max | Alibaba (Qwen) |
👁✓
82.3MMMU-Pro
🛠✓
—
🖥✓
86.1OSWorld-V
|
43.6S | 92.6S | — | — | — | 86.1S | 82.3S | — | $1.65 / $4.95 | 1M | — | |
| Grok 4.6 | xAI |
👁?
—
🛠✓
—
🖥?
—
|
— | — | — | — | — | — | — | — | $2 / $6 | 500K | — | |
| Claude Opus 5 | Anthropic |
👁✓
—
🛠✓
89.1TB 2.x
🖥✓
—
|
64.7S | — | — | 89.1D | — | — | — | — | $5 / $25 | 1M | — | |
| Kimi K3 | Moonshot (Kimi) |
👁✓
81.6MMMU-Pro
🛠✓
88.3TB 2.x
🖥✓
84.8OSWorld-V
|
56.0S | 93.5S | — | 88.3O | — | 84.8O | 81.6S | — | $2.85 / $14.3 | 1M | — | |
| Grok 4.5 | xAI |
👁?
—
🛠✓
83.3TB 2.x
🖥?
—
|
— | 93.0S | — | 83.3D | — | — | — | — | $2 / $6 | 500K | — | |
| Muse Spark 1.1 | Meta |
👁✓
—
🛠✓
—
🖥✓
80.8OSWorld-V
|
62.1S | — | — | — | — | 80.8S | — | — | $1.25 / $4.25 | 1M | — | |
| Gemini 3.6 Flash |
👁✓
—
🛠✓
—
🖥✓
83.0OSWorld-V
|
— | — | — | — | — | 83.0S | — | — | $1.5 / $7.5 | 1M | — | ||
| GPT-5.6 Luna | OpenAI |
👁✓
78.4MMMU-Pro
🛠✓
84.7TB 2.x
🖥✓
45.6OSWorld 2.0
|
— | 92.3S | — | 84.7D | — | — | 78.4S | — | $0.2 / $1.2 | 1.1M | — | |
| GPT-5.6 Terra | OpenAI |
👁✓
80.7MMMU-Pro
🛠✓
87.4TB 2.x
🖥✓
50.2OSWorld 2.0
|
— | 92.9S | — | 87.4D | — | — | 80.7S | — | $2 / $12 | 1.1M | — | |
| GPT-5.6 Sol | OpenAI |
👁✓
83.0MMMU-Pro
🛠✓
88.8TB 2.x
🖥✓
65.7OSWorld 2.0
|
— | 94.6S | — | 88.8D | — | — | 83.0S | — | $5 / $30 | 1.1M | — | |
| Claude Sonnet 5 | Anthropic |
👁✓
—
🛠✓
80.4TB 2.x
🖥✓
81.2OSWorld-V
|
57.4S | — | 85.2S | 80.4S | — | 81.2S | — | — | $2 / $10 | 1M | — | |
| Claude Fable 5 | Anthropic |
👁✓
87.3ScreenSpot-Pro
🛠✓
—
🖥✓
85.0OSWorld-V
|
64.5S | — | 95.0S | — | — | 85.0S | — | — | $10 / $50 | 1M | — | |
| Nemotron 3 Ultra (550B) | NVIDIA |
👁?
—
🛠✓
—
🖥?
—
|
37.4S | — | — | — | — | — | — | — | $0.5 / $2.2 | 262K | Yes | |
| Claude Opus 4.8 | Anthropic |
👁✓
—
🛠✓
74.6TB 2.x
🖥✓
83.4OSWorld-V
|
57.9S | 93.6S | 88.6S | 74.6S | — | 83.4S | — | — | $5 / $25 | 1M | — | |
| Qwen3.7 Max | Alibaba (Qwen) |
👁?
—
🛠✓
69.7TB 2.x
🖥?
—
|
41.4S | 92.4S | 80.4S | 69.7S | — | — | — | — | $1.25 / $3.75 | 1M | — | |
| Grok 4.3 | xAI |
👁?
—
🛠✓
—
🖥?
—
|
— | — | — | — | — | — | — | — | $1.25 / $2.5 | 1M | — | |
| Gemini 3.5 Flash |
👁✓
83.6MMMU-Pro
🛠✓
76.2TB 2.x
🖥✓
78.4OSWorld-V
|
40.2S | — | — | 76.2S | — | 78.4S | 83.6S | — | $1.5 / $9 | 1M | — | ||
| Claude Opus 4.7 | Anthropic |
👁✓
—
🛠✓
69.4TB 2.x
🖥✓
78.0OSWorld-V
|
54.7S | 94.2S | 87.6S | 69.4S | — | 78.0S | — | — | $5 / $25 | 1M | — | |
| Claude Mythos Preview | Anthropic |
👁✓
—
🛠✓
82.0TB 2.x
🖥✓
79.6OSWorld-V
|
64.7S | 94.6S | 93.9S | 82.0S | — | 79.6S | — | — | — | — | — | |
| Mistral Medium 3.5 | Mistral AI |
👁?
—
🛠✓
—
🖥?
—
|
— | — | 77.6S | — | — | — | — | — | $1.5 / $7.5 | 256K | — | |
| Kimi K2.6 | Moonshot (Kimi) |
👁✓
80.1MMMU-Pro
🛠✓
66.7TB 2.x
🖥✓
73.1OSWorld-V
|
36.4S | 90.5S | 80.2S | 66.7S | — | 73.1S | 80.1S | — | $0.75 / $3.5 | 262K | — | |
| DeepSeek-V4-Flash-Max | DeepSeek |
👁?
—
🛠✓
56.9TB 2.x
🖥?
—
|
45.1S | 88.1S | 79.0S | 56.9S | — | — | — | — | $0.09 / $0.18 | 1M | — | |
| DeepSeek-V4-Pro-Max | DeepSeek |
👁?
—
🛠✓
67.9TB 2.x
🖥?
—
|
48.2S | 90.1S | 80.6S | 67.9S | — | — | — | — | $1.3 / $2.6 | 1M | — | |
| Muse Spark | Meta |
👁✓
80.4MMMU-Pro
🛠✓
59.0TB 2.x
🖥?
—
|
58.4S | 89.5S | 77.4S | 59.0S | 91.5S | — | 80.4S | — | — | — | — | |
| Gemma 4 31B |
👁✓
76.9MMMU-Pro
🛠✓
—
🖥?
—
|
— | 84.3O | — | — | — | — | 76.9S | — | $0.09 / $0.34 | 262K | Yes | ||
| GPT-5.5 | OpenAI |
👁✓
83.2MMMU-Pro
🛠✓
82.7TB 2.x
🖥✓
78.7OSWorld-V
|
52.2S | 93.6S | — | 82.7S | 98.0S | 78.7S | 83.2S | — | $5 / $30 | 1.1M | — | |
| Nemotron 3 Super (120B) | NVIDIA |
👁?
—
🛠✓
64.4τ²-Bench
🖥?
—
|
— | 79.2O | — | — | 64.4S | — | — | — | $0.09 / $0.4 | 262K | Yes | |
| Mistral Small 4 | Mistral AI |
👁?
—
🛠✓
—
🖥?
—
|
— | — | — | — | — | — | — | — | $0.15 / $0.6 | 256K | — | |
| Qwen3.6 Plus | Alibaba (Qwen) |
👁✓
78.8MMMU-Pro
🛠✓
61.6TB 2.x
🖥✓
62.5OSWorld-V
|
— | 90.4S | 78.8S | 61.6S | — | 62.5S | 78.8S | 86.0S | $0.5 / $3 | 1M | — | |
| GPT-5.4 mini | OpenAI |
👁✓
76.6MMMU-Pro
🛠✓
60.0TB 2.x
🖥✓
72.1OSWorld-V
|
— | 88.0S | — | 60.0S | 93.4S | 72.1S | 76.6S | — | $0.75 / $4.5 | 400K | — | |
| GPT-5.4 | OpenAI |
👁✓
81.2MMMU-Pro
🛠✓
75.1TB 2.x
🖥✓
75.0OSWorld-V
|
39.8S | 92.8S | — | 75.1S | 98.9S | 75.0S | 81.2S | — | $2.5 / $15 | 1M | — | |
| Claude Sonnet 4.6 | Anthropic |
👁✓
75.6MMMU-Pro
🛠✓
59.1TB 2.x
🖥✓
78.5OSWorld-V
|
49.0S | 89.9S | 79.6S | 59.1S | 97.9S | 78.5D | 75.6S | — | $3 / $15 | 1M | — | |
| Claude Opus 4.6 | Anthropic |
👁✓
77.3MMMU-Pro
🛠✓
65.4TB 2.x
🖥✓
—
|
53.1S | 91.3S | 80.8S | 65.4S | 99.3S | — | 77.3S | — | $5 / $25 | 1M | — | |
| Qwen3.5-397B-A17B | Alibaba (Qwen) |
👁?
—
🛠✓
52.5TB 2.x
🖥?
—
|
— | 88.4S | 76.4S | 52.5S | — | — | — | — | $0.45 / $3 | 262K | Yes | |
| Gemini 3.1 Pro |
👁✓
80.5MMMU-Pro
🛠✓
68.5TB 2.x
🖥?
—
|
51.4S | 94.3S | 80.6S | 68.5S | — | — | 80.5S | — | $2 / $12 | 1M | — | ||
| GPT-5.3 Codex | OpenAI |
👁✓
—
🛠✓
77.3TB 2.x
🖥✓
64.7OSWorld-V
|
— | — | — | 77.3S | — | 64.7S | — | — | $1.75 / $14 | 400K | — | |
| Kimi K2.5 | Moonshot (Kimi) |
👁✓
78.5MMMU-Pro
🛠✓
50.8TB 2.x
🖥?
—
|
50.2S | 87.6S | 76.8S | 50.8S | — | — | 78.5S | — | — | — | — | |
| Nemotron 3 Nano (30B) | NVIDIA |
👁✕
—
🛠✓
42.2τ²-Bench
🖥✕
—
|
— | — | — | — | 42.2S | — | — | — | $0.05 / $0.2 | 262K | Yes | |
| Mistral Large 3 | Mistral AI |
👁?
—
🛠✓
—
🖥?
—
|
— | — | — | — | — | — | — | — | $0.5 / $1.5 | 262K | Yes | |
| Qwen3 Max | Alibaba (Qwen) |
👁?
—
🛠✓
—
🖥?
—
|
— | — | — | — | — | — | — | — | $1.2 / $6 | 256K | — | |
| DeepSeek-V3.2 | DeepSeek |
👁✕
—
🛠✓
46.4TB 2.x
🖥✕
—
|
40.8S | 82.4O | — | 46.4S | — | — | — | — | $0.26 / $0.38 | 164K | Yes | |
| Gemini 3 Flash |
👁✓
81.2MMMU-Pro
🛠✓
47.6TB 2.x
🖥?
—
|
43.5S | 90.4S | 78.0S | 47.6S | — | — | 81.2S | — | $0.5 / $3 | 1M | — | ||
| GPT-5.2 | OpenAI |
👁✓
79.5MMMU-Pro
🛠✓
98.7τ²-Bench
🖥?
—
|
34.5S | 92.4S | 80.0S | — | 98.7S | — | 79.5S | — | $1.75 / $14 | 400K | — | |
| Claude Opus 4.5 | Anthropic |
👁✓
—
🛠✓
59.3TB 2.x
🖥✓
—
|
— | 87.0O | 80.9S | 59.3S | 98.2S | — | — | — | $5 / $25 | 200K | — | |
| Grok-4.1 Fast | xAI |
👁?
—
🛠✓
—
🖥?
—
|
— | — | — | — | — | — | — | — | $0.2 / $0.5 | 2M | — | |
| Gemini 3 Pro |
👁✓
81.0MMMU-Pro
🛠✓
54.2TB 2.x
🖥?
—
|
45.8S | 91.9S | 76.2S | 54.2S | — | — | 81.0S | — | — | — | — | ||
| GPT-5.1 | OpenAI |
👁✓
85.4MMMU
🛠✓
95.6τ²-Bench
🖥?
—
|
— | 88.1S | 76.3S | — | 95.6S | — | — | 85.4S | $1.25 / $10 | 400K | — | |
| Claude Haiku 4.5 | Anthropic |
👁✓
—
🛠✓
41.8TB 2.x
🖥✓
—
|
— | 73.0O | 73.3S | 41.8O | 83.0S | — | — | — | $1 / $5 | 200K | — | |
| Claude Sonnet 4.5 | Anthropic |
👁✓
—
🛠✓
—
🖥✓
61.4OSWorld*
|
— | 83.4O | 77.2O | — | — | 61.4*O | — | — | $3 / $15 | 200K | — | |
| Claude Opus 4.1 | Anthropic |
👁✓
—
🛠✓
—
🖥✓
—
|
— | 80.9O | 74.5S | — | — | — | — | — | $15 / $75 | 200K | — | |
| Nemotron Nano 9B v2 | NVIDIA |
👁✕
—
🛠✓
—
🖥✕
—
|
— | — | — | — | — | — | — | — | — | — | Yes | |
| GPT-5 | OpenAI |
👁✓
78.4MMMU-Pro
🛠✓
96.7τ²-Bench
🖥?
—
|
— | — | 74.9S | — | 96.7S | — | 78.4S | 84.2S | — | — | — | |
| GPT OSS 120B | OpenAI |
👁✕
—
🛠✓
—
🖥✕
—
|
— | 80.1O | 62.4O | — | — | — | — | — | $0.04 / $0.17 | 131K | Yes | |
| Kimi K2 | Moonshot (Kimi) |
👁✕
—
🛠✓
65.8τ²-Bench
🖥✕
—
|
— | 75.1O | 65.8O | — | 65.8S | — | — | — | — | — | Yes | |
| Qwen3-Coder 480B | Alibaba (Qwen) |
👁✕
—
🛠✓
37.5TB 2.x
🖥✕
—
|
— | — | 67.0O | 37.5S | — | — | — | — | $0.3 / $1 | 262K | Yes | |
| Grok-4 Heavy | xAI |
👁?
—
🛠✓
—
🖥?
—
|
50.7S | 88.4S | — | — | — | — | — | — | — | — | — | |
| Grok-4 | xAI |
👁?
—
🛠✓
—
🖥?
—
|
40.0S | 87.5S | — | — | — | — | — | — | — | — | — | |
| Claude Sonnet 4 | Anthropic |
👁✓
74.4MMMU
🛠✓
—
🖥✓
42.2OSWorld*
|
— | — | 72.7O | — | — | 42.2*O | — | 74.4S | $3 / $15 | 200K | — | |
| Claude Opus 4 | Anthropic |
👁✓
—
🛠✓
—
🖥✓
—
|
— | — | 72.5O | — | — | — | — | — | $15 / $75 | 200K | — | |
| Gemini 2.5 Flash |
👁✓
79.7MMMU
🛠✓
—
🖥?
—
|
— | 82.8O | 60.4O | — | — | — | — | 79.7S | $0.3 / $2.5 | 1M | — | ||
| Gemini 2.5 Pro |
👁✓
79.6MMMU
🛠✓
—
🖥?
—
|
— | 84.0O | 63.8O | — | — | — | — | 79.6S | $1.25 / $10 | 1M | — | ||
| Llama 3.1 Nemotron Ultra 253B | NVIDIA |
👁✕
—
🛠✓
—
🖥✕
—
|
— | 72.8O | — | — | — | — | — | — | — | 128K | Yes | |
| Qwen3 32B | Alibaba (Qwen) |
👁✕
—
🛠✓
—
🖥✕
—
|
— | — | — | — | — | — | — | — | $0.08 / $0.28 | 128K | Yes | |
| Llama 4 Maverick | Meta |
👁✓
73.4MMMU
🛠✓
—
🖥?
—
|
— | 69.8O | — | — | — | — | — | 73.4S | $0.2 / $0.8 | 1M | Yes | |
| o4-mini | OpenAI |
👁✓
81.6MMMU
🛠✓
—
🖥?
—
|
— | 81.4O | 68.1O | — | — | — | — | 81.6S | $1.1 / $4.4 | 200K | — | |
| o3 | OpenAI |
👁✓
76.4MMMU-Pro
🛠✓
58.2τ²-Bench
🖥?
—
|
— | 83.3O | 69.1O | — | 58.2S | — | 76.4S | 82.9S | — | — | — | |
| GPT-4.1 | OpenAI |
👁✓
74.8MMMU
🛠✓
—
🖥?
—
|
— | 66.3O | 54.6O | — | — | — | — | 74.8S | $2 / $8 | 1M | — | |
| Gemma 3 27B |
👁?
—
🛠✓
—
🖥?
—
|
— | 42.4O | — | — | — | — | — | — | $0.08 / $0.16 | 131K | Yes | ||
| Claude 3.7 Sonnet | Anthropic |
👁✓
75.0MMMU
🛠✓
—
🖥✓
—
|
— | 84.8O | 63.7O | — | — | — | — | 75.0S | $3 / $15 | 200K | — | |
| Grok-3 | xAI |
👁✓
78.0MMMU
🛠✓
—
🖥?
—
|
— | 84.6O | — | — | — | — | — | 78.0S | $3 / $15 | 128K | — | |
| GPT-4.5 | OpenAI |
👁✓
75.2MMMU
🛠✓
—
🖥?
—
|
— | 71.4O | 38.0O | — | — | — | — | 75.2S | — | — | — | |
| Mistral Small 3 | Mistral AI |
👁✕
—
🛠✓
—
🖥✕
—
|
— | — | — | — | — | — | — | — | $0.05 / $0.08 | 33K | Yes | |
| DeepSeek-R1 | DeepSeek |
👁✕
—
🛠✕
—
🖥✕
—
|
— | 71.5O | 49.2O | — | — | — | — | — | — | — | Yes | |
| o3-mini | OpenAI |
👁✕
—
🛠✓
—
🖥✕
—
|
— | 79.7O | 49.3O | — | — | — | — | — | $1.1 / $4.4 | 200K | — | |
| DeepSeek-V3 | DeepSeek |
👁✕
—
🛠✓
—
🖥✕
—
|
— | 59.1O | 42.0O | — | — | — | — | — | $0.32 / $0.89 | 164K | Yes | |
| Llama 3.3 70B | Meta |
👁✕
—
🛠✓
—
🖥✕
—
|
— | 50.5O | — | — | — | — | — | — | $0.1 / $0.32 | 131K | Yes | |
| Gemini 2.0 Flash |
👁✓
70.7MMMU
🛠✓
—
🖥?
—
|
— | 65.2O | 21.4O | — | — | — | — | 70.7S | — | — | — | ||
| o1 | OpenAI |
👁✓
77.6MMMU
🛠✓
—
🖥?
—
|
— | 78.0O | 48.9O | — | — | — | — | 77.6S | — | — | — | |
| Claude 3.5 Haiku | Anthropic |
👁✓
—
🛠✓
—
🖥?
—
|
— | — | 40.6O | — | — | — | — | — | $0.8 / $4 | 200K | — | |
| Claude 3.5 Sonnet (new) | Anthropic |
👁✓
—
🛠✓
—
🖥✓
14.9OSWorld*
|
— | — | 49.0O | — | — | 14.9*O | — | — | $3 / $15 | 200K | — | |
| Llama 3.1 Nemotron 70B | NVIDIA |
👁✕
—
🛠✓
—
🖥✕
—
|
— | — | — | — | — | — | — | — | — | — | Yes | |
| Qwen2.5 72B | Alibaba (Qwen) |
👁✕
—
🛠✓
—
🖥✕
—
|
— | 49.0O | — | — | — | — | — | — | $0.36 / $0.4 | 33K | Yes | |
| Grok-2 | xAI |
👁✓
66.1MMMU
🛠✓
—
🖥✕
—
|
— | 56.0O | — | — | — | — | — | 66.1S | — | — | — | |
| Mistral Large 2 | Mistral AI |
👁✕
—
🛠✓
—
🖥✕
—
|
— | — | — | — | — | — | — | — | — | — | — | |
| Llama 3.1 405B | Meta |
👁✕
—
🛠✓
—
🖥✕
—
|
— | 50.7O | — | — | — | — | — | — | — | — | Yes | |
| GPT-4o mini | OpenAI |
👁✓
59.4MMMU
🛠✓
—
🖥✕
—
|
— | 40.2O | — | — | — | — | — | 59.4S | $0.15 / $0.6 | 128K | — | |
| Claude 3.5 Sonnet | Anthropic |
👁✓
68.3MMMU
🛠✓
—
🖥✕
—
|
— | 59.4O | 33.4O | — | — | — | — | 68.3S | $3 / $15 | 200K | — | |
| Codestral 22B | Mistral AI |
👁✕
—
🛠?
—
🖥✕
—
|
— | — | — | — | — | — | — | — | — | — | — | |
| DeepSeek-V2.5 | DeepSeek |
👁✕
—
🛠?
—
🖥✕
—
|
— | — | — | — | — | — | — | — | — | — | Yes | |
| Gemini 1.5 Flash |
👁✓
62.3MMMU
🛠?
—
🖥✕
—
|
— | — | — | — | — | — | — | 62.3S | — | — | — | ||
| Gemini 1.5 Pro |
👁✓
65.9MMMU
🛠?
—
🖥✕
—
|
— | — | — | — | — | — | — | 65.9S | — | — | — | ||
| GPT-4o | OpenAI |
👁✓
72.2MMMU
🛠✓
23.5τ²-Bench
🖥✕
—
|
— | 53.6O | 33.2O | — | 23.5S | — | — | 72.2S | $2.5 / $10 | 128K | — | |
| GPT-4 Turbo | OpenAI |
👁?
—
🛠?
—
🖥✕
—
|
— | — | — | — | — | — | — | — | $10 / $30 | 128K | — | |
| Claude 3 Haiku | Anthropic |
👁✓
—
🛠?
—
🖥✕
—
|
— | 33.3O | — | — | — | — | — | — | $0.25 / $1.25 | 200K | — | |
| Claude 3 Sonnet | Anthropic |
👁✓
—
🛠?
—
🖥✕
—
|
— | 40.4O | — | — | — | — | — | — | $3 / $15 | 200K | — | |
| Claude 3 Opus | Anthropic |
👁✓
—
🛠?
—
🖥✕
—
|
— | 50.4O | — | — | — | — | — | — | $15 / $75 | 200K | — | |
| Grok-1.5 | xAI |
👁✕
—
🛠✕
—
🖥✕
—
|
— | — | — | — | — | — | — | — | — | — | — | |
| Gemini 1.0 Pro |
👁✓
—
🛠?
—
🖥✕
—
|
— | — | — | — | — | — | — | — | — | — | — | ||
| Claude 2.1 | Anthropic |
👁✕
—
🛠?
—
🖥✕
—
|
— | — | — | — | — | — | — | — | — | 200K | — | |
| Claude 2 | Anthropic |
👁✕
—
🛠✕
—
🖥✕
—
|
— | — | — | — | — | — | — | — | — | 100K | — | |
| Claude | Anthropic |
👁✕
—
🛠✕
—
🖥✕
—
|
— | — | — | — | — | — | — | — | — | — | — | |
| GPT-4 | OpenAI |
👁✕
—
🛠?
—
🖥✕
—
|
— | — | — | — | — | — | — | — | — | 8K | — | |
| GPT-3.5 Turbo | OpenAI |
👁✕
—
🛠?
—
🖥✕
—
|
— | — | — | — | — | — | — | — | $0.5 / $1.5 | 16K | — |
✓ yes · ✕ no · ? no clear info
* Original OSWorld (pre-Verified version) · O = official company announcement · S = llm-stats.com · D = independent compilation · — = No published score
What does each benchmark measure?
- Humanity's Last Exam
- The hardest expert-written academic questions
- GPQA Diamond
- PhD-level science questions
- SWE-bench Verified
- Fixing real GitHub issues (software)
- Terminal-Bench 2.x
- Agentic tasks in the terminal
- τ²-Bench (Telecom)
- Acting as an agent with tools to solve a customer issue end to end
- OSWorld-Verified
- Using a computer (screen, mouse, keyboard)
- MMMU-Pro
- Vision: multi-discipline questions with charts, diagrams and images (hard version)
- MMMU
- Vision: college-level questions with charts, diagrams and images
Method & sources
Dates, prices and context come from company docs and llm-stats.com organisation pages; for some models only the release month is known. Scores are either published in the company's official announcement (O), taken from llm-stats.com leaderboards (S), or compiled from independent write-ups of the company's launch tables (D). For the newest models companies report newer benchmark versions (e.g. OSWorld 2.0, Terminal-Bench 4.0); these appear in the capability column with their version name and should not be compared directly with older versions. Sources may use different settings, so treat scores as directional. No unpublished value was estimated. The table lists each company's notable models, not every sub-variant.
- OAnthropic — duyurular ve Models overview
- OOpenAI — model duyuruları
- OGoogle DeepMind — Gemini
- OMeta AI — Llama
- ODeepSeek — model raporları
- SLLM Stats — SWE-bench Verified
- SLLM Stats — GPQA Diamond
- SLLM Stats — Terminal-Bench 2.0
- SLLM Stats — OSWorld-Verified
- SLLM Stats — Humanity's Last Exam
- SLLM Stats — MMMU / MMMU-Pro
- SLLM Stats — τ²-Bench Telecom
- SLLM Stats — şirket sayfaları (tarih, fiyat, bağlam)
- DVellum — GPT-6 Astra benchmark açıklaması
- DVellum — Claude Sonnet 5 benchmark açıklaması
- DOfficeChai — GPT-6 Astra resmi benchmarkları
- DApidog — Grok 4.5 yayımlanan skorlar
- DMindStudio — DeepSeek V4 Pro 0813 / Kimi K3 karşılaştırması
- OMoonshot — Kimi K3 model kartı
- DVellum — GPT-5.6 Sol / Terra / Luna
- DVellum — Gemini 3.8 Flash benchmark açıklaması
- DApidog — DeepSeek-V4.1-Flash görsel API
- WWikipedia — Claude (language model)
🖼 Image generation models
Text-to-image and editing existing images. Scores come from LMArena votes where people compare two outputs without knowing which model made them.
77models
20companies
1421Text-to-image leader · GPT Image 2.5 Sunburst
18open-weight
Top 10
Each company's best model
- OpenAIGPT Image 2.5 Sunburst · Sept 20261520
- xAIGrok Imagine Image 2.0 (low) · Aug 20261439
- Microsoft AIMAI Image 2.6 · 14 Aug 20261434
- MetaMuse Image · 7 Jul 20261403
- ByteDanceSeedream 5.0 Pro · 8 Jul 20261394
- GoogleGemini 3 Pro Image 2k (nano Banana Pro) · Nov 20251390
- ReveReve 2.1 · 9 Jul 20261375
- Luma AIUni 1.1 Max · May 20261333
- AlibabaQwen Image 2.0 Pro 2026 06 22 · 22 Jun 20261304
- Black Forest LabsFlux 2 Max · 25 Nov 20251262
- IdeogramIdeogram 4.0 Quality · 3 Jun 20261204
- RecraftRecraft v4.1 Utility Pro · 13 May 20261169
- NVIDIACosmos3 Super Text2Image (Agentic) · Jun 20261167
- TencentHunyuan Image 3.0 · Sept 20251151
- KreaKrea 2 Medium · 12 May 20261123
- HiDreamHidream O1 Image · 8 May 20261117
- RunwayRunway Gen4 · May 20251025
- Leonardo AILucid Origin · Aug 20251013
- Z.aiGLM Image · Jan 20261010
- Stability AIStable Diffusion v35 Large · Oct 2024938
All models
77 models shown
| Model | Company | Released | Text-to-image (Elo) | Image editing (Elo) | Open | Votes | License |
|---|---|---|---|---|---|---|---|
| GPT Image 2.5 Sunburst gpt-image-2.5-sunburst |
OpenAI | 1421 | 1520 | No | 3,149 | Proprietary | |
| GPT Image 2.5 Flare gpt-image-2.5-flare |
OpenAI | 1399 | 1491 | No | 2,856 | Proprietary | |
| GPT Image 2 (medium) gpt-image-2 (medium) |
OpenAI | 1381 | 1461 | No | 78,731 | Proprietary | |
| MAI Image 2.6 mai-image-2.6 |
Microsoft AI | 1331 | 1434 | No | 11,213 | Proprietary | |
| Grok Imagine Image 2.0 (low) grok-imagine-image-2.0 (low) |
xAI | 1315 | 1439 | No | 2,681 | Proprietary | |
| Reve 2.1 reve-2.1 |
Reve | 1301 | 1375 | No | 7,573 | Proprietary | |
| Muse Image muse-image |
Meta | 1277 | 1403 | No | 26,045 | Proprietary | |
| Reve 2.0 reve-2.0 |
Reve | 1270 | 1358 | No | 14,632 | Proprietary | |
| Gemini 3.1 Flash Image (nano Banana 2) gemini-3.1-flash-image (nano-banana-2) |
1261 | 1387 | No | 41,957 | Proprietary | ||
| Seedream 5.0 Pro seedream-5.0-pro |
ByteDance | 1257 | 1394 | No | 62,443 | Proprietary | |
| Qwen Image 3.0 Pro qwen-image-3.0-pro |
Alibaba | 1254 | — | No | 10,920 | Proprietary | |
| MAI Image 2.5 mai-image-2.5 |
Microsoft AI | 1254 | 1400 | No | 57,250 | Proprietary | |
| Gemini 3.1 Flash Lite Image gemini-3.1-flash-lite-image |
— | 1250 | 1314 | No | 16,371 | Proprietary | |
| Gemini 3 Pro Image 2k (nano Banana Pro) gemini-3-pro-image-2k (nano-banana-pro) |
1246 | 1390 | No | 152,220 | Proprietary | ||
| GPT Image 1.5 High Fidelity gpt-image-1.5-high-fidelity |
OpenAI | 1239 | 1370 | No | 151,702 | Proprietary | |
| Gemini 3 Pro Image Preview gemini-3-pro-image-preview |
1232 | 1385 | No | 82,737 | Proprietary | ||
| Ideogram 4.0 Quality ideogram-4.0-quality |
Ideogram | 1204 | — | 🔓 Yes | 42,169 | Ideogram Open Model | |
| Qwen Image 2.0 Pro 2026 06 22 qwen-image-2.0-pro-2026-06-22 |
Alibaba | 1191 | 1304 | No | 12,005 | Proprietary | |
| Uni 1.1 Max uni-1.1-max |
Luma AI | 1188 | 1333 | No | 13,469 | Proprietary | |
| MAI Image 2 mai-image-2 |
Microsoft AI | 1183 | — | No | 49,198 | Proprietary | |
| Uni 1.1 uni-1.1 |
Luma AI | 1181 | 1314 | No | 33,836 | Proprietary | |
| Grok Imagine Image grok-imagine-image |
xAI | 1171 | 1330 | No | 236,623 | Proprietary | |
| Recraft v4.1 Utility Pro recraft-v4.1-utility-pro |
Recraft | 1169 | — | No | 2,516 | Proprietary | |
| Cosmos3 Super Text2Image (Agentic) Cosmos3-Super-Text2Image (Agentic) |
NVIDIA | 1167 | — | 🔓 Yes | 7,424 | OpenMDW-1.1 | |
| Flux 2 Max flux-2-max |
Black Forest Labs | 1162 | 1262 | No | 117,339 | Proprietary | |
| Grok Imagine Image Pro grok-imagine-image-pro |
xAI | 1161 | — | No | 93,639 | Proprietary | |
| Flux 2 Flex flux-2-flex |
Black Forest Labs | 1157 | 1225 | No | 149,288 | Proprietary | |
| Flux 2 Pro flux-2-pro |
Black Forest Labs | 1154 | 1245 | No | 188,083 | Proprietary | |
| Reve v1.5 reve-v1.5 |
Reve | — | 1154 | — | No | 36,748 | Proprietary |
| Cosmos3 Super Text2Image Cosmos3-Super-Text2Image |
NVIDIA | 1154 | — | 🔓 Yes | 12,499 | OpenMDW-1.1 | |
| Hunyuan Image 3.0 hunyuan-image-3.0 |
Tencent | 1151 | — | 🔓 Yes | 173,099 | tencent-hunyuan-community | |
| Gemini 2.5 Flash Image Preview (nano Banana) gemini-2.5-flash-image-preview (nano-banana) |
1150 | 1293 | No | 845,086 | Proprietary | ||
| Imagen Ultra 4.0 Generate 001 imagen-ultra-4.0-generate-001 |
1148 | — | No | 388,103 | Proprietary | ||
| Seedream 4.5 seedream-4.5 |
ByteDance | 1147 | 1302 | No | 259,145 | Proprietary | |
| Flux 2 Dev flux-2-dev |
Black Forest Labs | 1145 | 1226 | 🔓 Yes | 73,963 | flux-non-commercial-license | |
| Seedream 4 2K seedream-4-2k |
ByteDance | 1140 | 1271 | No | 12,621 | Proprietary | |
| Seedream 5.0 Lite seedream-5.0-lite |
ByteDance | 1138 | 1294 | No | 109,182 | Proprietary | |
| Wan2.6 T2I wan2.6-t2i |
Alibaba | 1137 | — | No | 201,108 | Proprietary | |
| Recraft v4.1 Pro recraft-v4.1-pro |
Recraft | 1130 | — | No | 2,680 | Proprietary | |
| Imagen 4.0 Generate 001 imagen-4.0-generate-001 |
1129 | — | No | 531,811 | Proprietary | ||
| Qwen Image 2512 qwen-image-2512 |
Alibaba | 1125 | — | 🔓 Yes | 92,541 | Apache 2.0 | |
| Krea 2 Medium krea-2-medium |
Krea | 1123 | — | No | 40,205 | Proprietary | |
| Wan2.5 T2I Preview wan2.5-t2i-preview |
Alibaba | 1118 | — | No | 248,411 | Proprietary | |
| Hidream O1 Image hidream-o1-image |
HiDream | 1117 | — | 🔓 Yes | 48,201 | MIT | |
| Seedream 4 Fal seedream-4-fal |
ByteDance | 1116 | 1210 | No | 11,870 | Proprietary | |
| GPT Image 1 gpt-image-1 |
OpenAI | 1115 | 1139 | No | 264,897 | Proprietary | |
| Recraft v4 recraft-v4 |
Recraft | 1114 | — | No | 114,295 | Proprietary | |
| Seedream 4 High Res Fal seedream-4-high-res-fal |
ByteDance | 1113 | 1217 | No | 175,361 | Proprietary | |
| Krea 2 Turbo krea-2-turbo |
Krea | 1110 | — | 🔓 Yes | 33,237 | krea-2-community-license | |
| GPT Image 1 Mini gpt-image-1-mini |
OpenAI | 1109 | 1124 | No | 163,843 | Proprietary | |
| Krea 2 Large krea-2-large |
Krea | 1108 | — | No | 39,614 | Proprietary | |
| Wan2.7 Image Pro wan2.7-image-pro |
Alibaba | 1103 | 1302 | No | 28,729 | Proprietary | |
| Wan2.7 Image wan2.7-image |
Alibaba | 1100 | 1301 | No | 29,089 | Proprietary | |
| MAI Image 1 mai-image-1 |
Microsoft AI | 1093 | — | No | 94,562 | Proprietary | |
| Z Image Turbo z-image-turbo |
Alibaba | 1084 | — | 🔓 Yes | 22,251 | Apache 2.0 | |
| Seedream 3 seedream-3 |
ByteDance | 1082 | — | No | 36,920 | Proprietary | |
| Flux 1 Kontext Max flux-1-kontext-max |
Black Forest Labs | 1074 | 1181 | No | 65,618 | Proprietary | |
| Flux 2 Klein 9b flux-2-klein-9b |
Black Forest Labs | 1070 | 1224 | 🔓 Yes | 142,642 | flux-non-commercial-license | |
| Qwen Image Prompt Extend qwen-image-prompt-extend |
Alibaba | 1061 | — | 🔓 Yes | 705,193 | Apache 2.0 | |
| Flux 1 Kontext Pro flux-1-kontext-pro |
Black Forest Labs | 1059 | 1176 | No | 330,851 | Proprietary | |
| Imagen 3.0 Generate 002 imagen-3.0-generate-002 |
1058 | — | No | 360,152 | Proprietary | ||
| Qwen Image qwen-image |
Alibaba | 1057 | — | 🔓 Yes | 84,717 | Apache 2.0 | |
| Ideogram v3 Quality ideogram-v3-quality |
Ideogram | 1049 | — | No | 115,413 | Proprietary | |
| Photon photon |
Luma AI | 1035 | — | No | 127,530 | Proprietary | |
| Flux 2 Klein 4b flux-2-klein-4b |
Black Forest Labs | 1030 | 1188 | 🔓 Yes | 144,483 | Apache 2.0 | |
| Runway Gen4 runway-gen4 |
Runway | 1025 | — | No | 51,755 | Proprietary | |
| Recraft v3 recraft-v3 |
Recraft | 1021 | — | No | 191,836 | Proprietary | |
| Flux 1.1 Pro flux-1.1-pro |
Black Forest Labs | 1016 | — | No | 70,460 | Proprietary | |
| Lucid Origin lucid-origin |
Leonardo AI | 1013 | — | No | 285,042 | Proprietary | |
| Ideogram v2 ideogram-v2 |
Ideogram | 1013 | — | No | 72,090 | Proprietary | |
| GLM Image glm-image |
Z.ai | 1010 | — | 🔓 Yes | 4,624 | MIT | |
| Gemini 2.0 Flash Preview Image Generation gemini-2.0-flash-preview-image-generation |
975 | 1081 | No | 257,085 | Proprietary | ||
| Flux 1 Dev FP8 flux-1-dev-fp8 |
Black Forest Labs | 969 | — | 🔓 Yes | 49,221 | Open | |
| DALL·E 3 dall-e-3 |
OpenAI | 968 | — | No | 239,060 | Proprietary | |
| Flux 1 Kontext Dev flux-1-kontext-dev |
Black Forest Labs | 940 | 1149 | 🔓 Yes | 215,336 | flux-1-dev-non-commercial-license | |
| Stable Diffusion v35 Large stable-diffusion-v35-large |
Stability AI | 938 | — | 🔓 Yes | 23,396 | Open | |
| Bagel bagel |
ByteDance | — | 898 | 1026 | 🔓 Yes | 12,375 | Apache 2.0 |
What is Elo?
Elo works like chess ratings: a model gains points when it wins votes against rivals. A 100-point gap means the higher model is preferred about 64% of the time. New models with few votes have higher uncertainty.
Different settings of the same model (resolution, fast variant, with/without audio) appear as separate rows. Some popular tools without a public API, such as Midjourney, are not in these votes.
Sources
- LMArenaLMArena — Text-to-Image
- LMArenaLMArena — Image Edit
Data as of 17 September 2026. Audio is marked only where the company announced it; others show “?”.
🎬 Video generation models
Text-to-video and image-to-video: Seedance, Veo, Sora, Kling, Hailuo, Wan, Runway and more; scores come from LMArena blind votes.
56models
19companies
1515Text-to-video leader · Gemini Omni 1.1 Flash
10generate their own audio
Top 10
Each company's best model
- GoogleGemini Omni 1.1 Flash · Aug 20261515
- AlibabaWan3.0 · 6 Aug 20261494
- Black Forest LabsFlux 3 Video · 23 Jul 20261494
- MiniMaxMinimax H3 · 31 Jul 20261494
- xAIGrok Imagine Video 1.5 Agent · Jun 20261491
- ByteDanceSeedance 2.5 720p · 31 Jul 20261482
- MetaMuse Video · 7 Jul 20261456
- HiDreamHidream O1 Video 1.01452
- OpenAISora 2 Pro · Sept 20251367
- Shengshu (Vidu)Vidu Q3 Pro · 30 Jan 20261363
- Kuaishou (Kling)Kling v3 Pro · 5 Feb 20261355
- PixVersePixverse v5.6 · 26 Jan 20261299
- Luma AIRay 3 · Sept 20251226
- RunwayRunway Gen 4.5 · Dec 20251224
- TencentHunyuan Video 1.5 · Nov 20251198
- KandinskyKandinsky 5.0 T2V Pro · Nov 20251172
- LightricksLTX 2 19b · Oct 20251159
- PikaPika v2.2 · Feb 20251009
- GenmoMochi v1 · Oct 20241006
All models
56 models shown
| Model | Company | Released | Text-to-video (Elo) | Image-to-video (Elo) | Audio | Open | Votes | License |
|---|---|---|---|---|---|---|---|---|
| Gemini Omni 1.1 Flash gemini-omni-1.1-flash |
1515 | 1488 | ? | No | 1,777 | Proprietary | ||
| Gemini Omni Flash gemini-omni-flash |
1511 | 1464 | ? | No | 21,934 | Proprietary | ||
| Wan3.0 wan3.0 |
Alibaba | 1494 | 1479 | ? | No | 1,167 | Proprietary | |
| Flux 3 Video flux-3-video |
Black Forest Labs | 1494 | 1450 | ? | No | 1,290 | Proprietary | |
| Grok Imagine Video 1.5 Agent grok-imagine-video-1.5-agent |
xAI | 1491 | 1456 | ? | No | 1,195 | Proprietary | |
| Seedance 2.5 720p dreamina-seedance-2.5-720p |
ByteDance | 1482 | 1475 | ? | No | 4,066 | Proprietary | |
| Seedance 2.0 720p dreamina-seedance-2.0-720p |
ByteDance | 1479 | 1474 | ? | No | 52,864 | Proprietary | |
| Minimax H3 minimax-h3 |
MiniMax | 1462 | 1494 | ? | 🔓 Yes | 7,648 | minimax-h3-community-license | |
| Muse Video muse-video |
Meta | 1456 | — | ? | No | 2,178 | Proprietary | |
| Hidream O1 Video 1.0 hidream-o1-video-1.0 |
HiDream | — | — | 1452 | ? | — | — | — |
| Happyhorse 1.0 happyhorse-1.0 |
Alibaba | 1427 | 1442 | ? | No | 22,116 | Proprietary | |
| Grok Imagine Video 480p grok-imagine-video-480p |
xAI | — | 1385 | ? | — | — | — | |
| Sora 2 Pro sora-2-pro |
OpenAI | 1367 | — | 🔊 Yes | No | 50,415 | Proprietary | |
| Veo 3.1 Audio veo-3.1-audio |
1364 | 1398 | 🔊 Yes | No | 13,703 | Proprietary | ||
| Veo 3.1 Audio 1080p veo-3.1-audio-1080p |
1363 | 1390 | 🔊 Yes | No | 24,979 | Proprietary | ||
| Vidu Q3 Pro vidu-q3-pro |
Shengshu (Vidu) | — | 1363 | ? | — | — | — | |
| Veo 3.1 Fast Audio veo-3.1-fast-audio |
1362 | 1385 | 🔊 Yes | No | 39,373 | Proprietary | ||
| Veo 3.1 Fast Audio 1080p veo-3.1-fast-audio-1080p |
1358 | 1372 | 🔊 Yes | No | 25,770 | Proprietary | ||
| Kling v3 Pro kling-v3-pro |
Kuaishou (Kling) | — | 1355 | ? | — | — | — | |
| Veo 3 Fast Audio veo-3-fast-audio |
1348 | 1326 | 🔊 Yes | No | 25,169 | Proprietary | ||
| Grok Imagine Video 720p grok-imagine-video-720p |
xAI | 1343 | 1415 | ? | No | 159,930 | Proprietary | |
| Sora 2 sora-2 |
OpenAI | 1342 | — | 🔊 Yes | No | 67,388 | Proprietary | |
| Wan2.7 T2V wan2.7-t2v |
Alibaba | 1341 | 1426 | ? | No | 23,025 | Proprietary | |
| Veo 3 Audio veo-3-audio |
1340 | 1331 | 🔊 Yes | No | 18,941 | Proprietary | ||
| Wan2.6 T2V wan2.6-t2v |
Alibaba | 1328 | 1310 | ? | No | 49,348 | Proprietary | |
| Seedance v1.5 Pro seedance-v1.5-pro |
ByteDance | 1256 | 1307 | ? | No | 75,891 | Proprietary | |
| Veo 3 veo-3 |
1253 | 1258 | ? | No | 14,951 | Proprietary | ||
| Veo 3 Fast veo-3-fast |
1248 | 1257 | ? | No | 15,225 | Proprietary | ||
| Wan2.5 T2V Preview wan2.5-t2v-preview |
Alibaba | 1246 | 1321 | ? | No | 32,461 | Proprietary | |
| Vidu Q2 Turbo vidu-q2-turbo |
Shengshu (Vidu) | — | 1243 | ? | — | — | — | |
| Pixverse v5.6 pixverse-v5.6 |
PixVerse | 1239 | 1299 | ? | No | 32,289 | Proprietary | |
| Kling v2.1 Standard kling-v2.1-standard |
Kuaishou (Kling) | — | 1229 | ? | — | — | — | |
| Runway Gen 4.5 runway-gen-4.5 |
Runway | 1224 | — | ? | No | 43,610 | Proprietary | |
| Vidu Q2 Pro vidu-q2-pro |
Shengshu (Vidu) | — | 1223 | ? | — | — | — | |
| Kling 2.5 Turbo 1080p kling-2.5-turbo-1080p |
Kuaishou (Kling) | 1219 | 1276 | No | No | 2,100 | Proprietary | |
| Kling 2.6 Pro kling-2.6-pro |
Kuaishou (Kling) | 1216 | 1294 | 🔊 Yes | No | 73,062 | Proprietary | |
| Ray 3 ray-3 |
Luma AI | 1205 | 1226 | ? | No | 1,129 | Proprietary | |
| Hailuo 2.3 hailuo-2.3 |
MiniMax | 1205 | 1262 | ? | No | 82,971 | Proprietary | |
| Kling O1 Pro kling-o1-pro |
Kuaishou (Kling) | 1205 | — | ? | No | 1,191 | Proprietary | |
| Hailuo 02 Pro hailuo-02-pro |
MiniMax | 1198 | 1229 | No | No | 9,363 | Proprietary | |
| Hailuo 02 Fast hailuo-02-fast |
MiniMax | — | 1194 | ? | — | — | — | |
| Seedance v1 Pro seedance-v1-pro |
ByteDance | 1190 | 1273 | ? | No | 12,120 | Proprietary | |
| Hailuo 02 Standard hailuo-02-standard |
MiniMax | 1181 | 1224 | No | No | 9,330 | Proprietary | |
| Kandinsky 5.0 T2V Pro kandinsky-5.0-t2v-pro |
Kandinsky | 1172 | — | ? | 🔓 Yes | 2,015 | MIT | |
| Hunyuan Video 1.5 hunyuan-video-1.5 |
Tencent | 1169 | 1198 | ? | 🔓 Yes | 4,278 | tencent-hunyuan-community | |
| Veo 2 veo-2 |
1164 | 1166 | No | No | 6,501 | Proprietary | ||
| Kling v2.1 Master kling-v2.1-master |
Kuaishou (Kling) | 1162 | 1235 | No | No | 14,048 | Proprietary | |
| LTX 2 19b ltx-2-19b |
Lightricks | 1154 | 1159 | 🔊 Yes | 🔓 Yes | 70,324 | ltx-2-community-license | |
| Wan v2.2 A14b wan-v2.2-a14b |
Alibaba | 1132 | 1170 | No | 🔓 Yes | 10,396 | Apache 2.0 | |
| Kandinsky 5.0 T2V Lite kandinsky-5.0-t2v-lite |
Kandinsky | 1113 | — | ? | 🔓 Yes | 1,468 | MIT | |
| Seedance v1 Lite seedance-v1-lite |
ByteDance | 1112 | 1185 | ? | No | 16,176 | Proprietary | |
| Sora sora |
OpenAI | 1069 | — | No | No | 4,069 | Proprietary | |
| Ray2 ray2 |
Luma AI | 1065 | 1108 | No | No | 5,222 | Proprietary | |
| Runway Gen4 Turbo runway-gen4-turbo |
Runway | — | 1052 | ? | — | — | — | |
| Pika v2.2 pika-v2.2 |
Pika | 1009 | 997 | No | No | 5,727 | Proprietary | |
| Mochi v1 mochi-v1 |
Genmo | 1006 | — | No | 🔓 Yes | 5,851 | Apache 2.0 |
What is Elo?
Elo works like chess ratings: a model gains points when it wins votes against rivals. A 100-point gap means the higher model is preferred about 64% of the time. New models with few votes have higher uncertainty.
Different settings of the same model (resolution, fast variant, with/without audio) appear as separate rows. Some popular tools without a public API, such as Midjourney, are not in these votes.
Sources
- LMArenaLMArena — Text-to-Video
- LMArenaLMArena — Image-to-Video
Data as of 17 September 2026. Audio is marked only where the company announced it; others show “?”.