AI model rankings
Benchmarks and reader votes answer different questions, so we keep them apart.
Your vote counts 1x. It rises to 2x after 7 days and 3x after 30.
| Model | Access | API price per 1M tokens | BenchLM | BenchLeader | Reader votes |
|---|---|---|---|---|---|
| Claude Opus 5.5Anthropic | Generally available | $4.00 in, $20.00 out | 86.3 | 69.2 | |
| Gemini 4 CarbonGoogle | Selected cyber defenders only | $2.00 in, $10.00 out | 88.4 | 68.9 | |
| GPT-6 AstraOpenAI | Generally available | $10.00 in, $50.00 out | 86.1 | 68.2 | |
| GPT-6.1 SolOpenAI | Generally available | $2.00 in, $10.00 out | 85.3 | 67.2 | |
| Claude Fable 5.1Anthropic | Generally available | $10.00 in, $50.00 out | 81.1 | 67.1 | |
| Claude Sonnet 5.5Anthropic | Generally available | $2.00 in, $10.00 out | 83.8 | 66.8 | |
| Qwen 3.8 MaxAlibaba Qwen | Generally available | $4.00 in, $12.00 out | 84.2 | 66.4 | |
| DeepSeek V4.1 FlashDeepSeek | Generally available | $0.04 in, $0.50 out | 83.5 | 66.1 | |
| Claude Fable 5Anthropic | Generally available | $10.00 in, $50.00 out | 80.8 | 66 | |
| Gemini 4 ArgonGoogle | Generally available | $2.00 in, $10.00 out | 82.3 | 65.9 | |
| Kimi K3.1 ProMoonshot AI | Generally available | $2.00 in, $10.00 out | 82 | 65.7 | |
| Claude Opus 5Anthropic | Generally available | $5.00 in, $25.00 out | 79.7 | 65.4 | |
| GPT-6.1 TerraOpenAI | Generally available | $1.00 in, $5.00 out | 80.5 | 65.2 | |
| Grok 4.7xAI | Generally available | $2.00 in, $6.00 out | 81.6 | 65 | |
| DeepSeek R1DeepSeek | Open Weights / API | $0.50 in, $2.15 out | 82.5 | 64.9 | |
| Mistral Large 4.0Mistral AI | Generally available | $0.70 in, $2.10 out | 81.2 | 64.8 | |
| Llama 4 MaverickMeta | Open Weights / API | $0.19 in, $0.65 out | 81 | 64.5 | |
| DeepSeek V3DeepSeek | Open Weights / API | $0.26 in, $1.03 out | 80.6 | 64.3 | |
| GPT-6 LunaOpenAI | Generally available | $0.10 in, $0.50 out | 79.9 | 64 | |
| Claude Haiku 5.5Anthropic | Generally available | $0.10 in, $0.50 out | 79.4 | 63.8 | |
| Gemini 3.8 FlashGoogle | Generally available | $0.83 in, $4.12 out | 79 | 63.5 | |
| Qwen 3.8 Omni FlashAlibaba Qwen | Generally available | $0.15 in, $0.47 out | 78.7 | 63.2 | |
| Gemini 2.5 ProGoogle | Generally available | $1.25 in, $10.00 out | 78.5 | 63.1 | |
| Muse Spark 1.3Meta | Generally available | $0.10 in, $0.20 out | 78.4 | 63 | |
| Command A+Cohere | Enterprise API | $0.30 in, $1.50 out | 78.2 | 62.8 | |
| MiniMax M3.1 FlashMiniMax | Generally available | $0.07 in, $0.30 out | 77.9 | 62.5 | |
| GPT-5 OmniOpenAI | Generally available | $1.25 in, $5.00 out | 77.8 | 62.4 | |
| Step-5 PreviewStepFun | Public Beta | $1.00 in, $2.70 out | 77.5 | 62.2 | |
| Grok 4.6xAI | Generally available | $2.00 in, $6.00 out | 77.3 | 62 | |
| Hunyuan 4 PreviewTencent | Public Beta | $0.83 in, $2.50 out | 77.1 | 61.9 | |
| Qwen 2.5 Coder 32BAlibaba Qwen | Open Weights / API | $0.66 in, $1.00 out | 77 | 61.8 | |
| GLM-5.3Z-AI (Zhipu) | Generally available | $0.07 in, $1.40 out | 76.8 | 61.6 | |
| Llama 4 ScoutMeta | Open Weights / API | $0.10 in, $0.30 out | 76.5 | 61.3 | |
| Kimi K3Moonshot AI | Generally available | $0.19 in, $15.00 out | 76.2 | 61 | |
| Mistral Medium 3.5Mistral AI | Generally available | $1.50 in, $7.50 out | 76 | 60.8 |
Run the race yourself
BenchLeader reports margins of about ±5 points on its top scores. Change the margin and see how often each model would come first.
Claude Opus 5.5 wins 0.0% of 20,000 simulated runs.
- Claude Opus 5.50.0%
- Gemini 4 Carbon0.0%
- GPT-6 Astra0.0%
- GPT-6.1 Sol0.0%
- Claude Fable 5.10.0%
- Claude Sonnet 5.50.0%
- Qwen 3.8 Max0.0%
- DeepSeek V4.1 Flash0.0%
- Claude Fable 50.0%
- Gemini 4 Argon0.0%
- Kimi K3.1 Pro0.0%
- Claude Opus 50.0%
- GPT-6.1 Terra0.0%
- Grok 4.70.0%
- DeepSeek R10.0%
- Mistral Large 4.00.0%
- Llama 4 Maverick0.0%
- DeepSeek V30.0%
- GPT-6 Luna0.0%
- Claude Haiku 5.50.0%
- Gemini 3.8 Flash0.0%
- Qwen 3.8 Omni Flash0.0%
- Gemini 2.5 Pro0.0%
- Muse Spark 1.30.0%
- Command A+0.0%
- MiniMax M3.1 Flash0.0%
- GPT-5 Omni0.0%
- Step-5 Preview0.0%
- Grok 4.60.0%
- Hunyuan 4 Preview0.0%
- Qwen 2.5 Coder 32B0.0%
- GLM-5.30.0%
- Llama 4 Scout0.0%
- Kimi K30.0%
- Mistral Medium 3.50.0%
See what a prompt costs
About 15 tokens. If you sent this a million times:
- GPT-6.1 Terra$15
- Claude Opus 5.5$60
Where the scores come from
BenchLM, as of 29 October 2026. A weighted average of eight benchmark categories. Gemini 4 Carbon's score is marked estimated.
BenchLeader Index, as of 31 October 2026. Combines published benchmarks. Top scores carry margins of about ±5.
Each board uses its own scale, so don't compare a BenchLM score with a BenchLeader score. Paid placement never affects rankings or votes.