Updated 22 October 2026
The best AI model right now
Claude Opus 5.5 from Anthropic is the strongest model on public benchmarks right now, with a capability score of 93.0.
| # | Model | Lab | Capability | Released |
|---|---|---|---|---|
| 1 | Claude Opus 5.5 | Anthropic | 93.0 | 2026-09 |
| 2 | Muse Wonder | Meta | 92.0 | 2026-10 |
| 3 | Gemini 4 Argon | Google DeepMind | 91.0 | 2026-09 |
| 4 | GPT-6 Astra | OpenAI | 86.0 | 2026-09 |
| 5 | Qwen3.8-Max | Alibaba Qwen | 80.0 | 2026-08 |
| 6 | Grok 4.7 | xAI | 80.0 | 2026-09 |
| 7 | Kimi K3 | Moonshot AI | 80.0 | 2026-07 |
| 8 | DeepSeek V4.1-Flash | DeepSeek | 79.0 | 2026-09 |
| 9 | Mistral Large 4 | Mistral AI | 69.0 | 2026-10 |
| 10 | Siri AI (iOS 27) | Apple | 55.0 | 2026-09 |
Capability is the average of each lab’s best model on the Artificial Analysis Intelligence Index and LMArena’s text leaderboard, scaled so the leader on each board scores 95. Which lab is winning also depends on release pace and money: see the live AI Race.
Who’s winning overall
| # | Lab | Index | 24h | Flagship |
|---|---|---|---|---|
| 1 | Meta | 94.3 | −0.1 | Muse Wonder |
| 2 | Anthropic | 91.6 | −0.1 | Claude Opus 5.5 |
| 3 | Google DeepMind | 89.7 | −0.1 | Gemini 4 Argon |
| 4 | Alibaba Qwen | 80.6 | −0.1 | Qwen3.8-Max |
| 5 | OpenAI | 80.4 | ±0.0 | GPT-6 Astra |
| 6 | DeepSeek | 73.8 | −0.1 | DeepSeek V4.1-Flash |
| 7 | xAI | 70.8 | −0.1 | Grok 4.7 |
| 8 | Moonshot AI | 58.8 | ±0.0 | Kimi K3 |
| 9 | Mistral AI | 56.5 | ±0.0 | Mistral Large 4 |
| 10 | Apple | 50.1 | ±0.0 | Siri AI (iOS 27) |