LiveBench Reasoning
LiveBench reasoning category.
- Published by
- LiveBench
- Category
- Reasoning
- Index weight
- Reference only
- Models
- 52
- Data as of
- 9 Sept 2026
LiveBench (livebench.ai), White et al. 2024.
- 1GPT-6 Astra92.7%
- 2Claude Fable 5.191.7%
- 3GPT-5.6 Sol91.7%
- 4Claude Opus 591.2%
- 5Kimi K390.7%
- 6GPT-5.6 Terra90.6%
- 7Grok 4.690.5%
- 8Muse Spark 1.290.0%
- 9Claude Fable 589.7%
- 10Muse Spark 1.389.7%
- 11GPT-5.589.7%
- 12Gemini 3.8 Flash89.3%
- 13Claude Opus 4.889.2%
- 14Claude Sonnet 588.7%
- 15Claude Opus 4.688.7%
52 of 52
| # | |||||
|---|---|---|---|---|---|
| 1 |  GPT-6 AstraOpenAI | 92.7% | – | 70.6 | – |
| 2 |  Claude Fable 5.1Anthropic | 91.7% | – | 70.2 | – |
| 3 | GPT-5.6 SolOpenAI | 91.7% | – | 65.5 | – |
| 4 | Claude Opus 5Anthropic | 91.2% | – | 66.8 | – |
| 5 |  Kimi K3Moonshot AIopen | 90.7% | kimi-k3 | 66.5 | – |
| 6 | GPT-5.6 TerraOpenAI | 90.6% | – | 58.0 | – |
| 7 |  Grok 4.6xAI | 90.5% | grok-4.6 | 63.6 | – |
| 8 |  Muse Spark 1.2Meta | 90.0% | – | 61.3 | – |
| 9 | Claude Fable 5Anthropic | 89.7% | – | 65.8 | – |
| 10 | Muse Spark 1.3Meta | 89.7% | – | 63.9 | – |
| 11 | GPT-5.5OpenAI | 89.7% | – | 63.7 | – |
| 12 |  Gemini 3.8 FlashGoogle | 89.3% | – | 60.1 | – |
| 13 | Claude Opus 4.8Anthropic | 89.2% | – | 59.4 | – |
| 14 | Claude Sonnet 5Anthropic | 88.7% | – | 57.3 | – |
| 15 | Claude Opus 4.6Anthropic | 88.7% | – | 61.3 | – |
| 16 |  Qwen3 8Alibaba | 88.2% | – | 63.9 | – |
| 17 | GPT-5.4OpenAI | 88.1% | – | 61.3 | – |
| 18 | Gemini 3.7 FlashGoogle | 87.8% | – | 60.8 | – |
| 19 | Muse Spark 1.1Meta | 87.7% | – | 56.4 | – |
| 20 | Qwen3.8-Flash-NextAlibabaopen | 87.4% | – | 59.3 | – |
| 21 | Claude Opus 4.7Anthropic | 87.2% | – | 57.2 | – |
| 22 | Grok 4.5xAI | 87.2% | grok-4.5 | 63.4 | – |
| 23 |  DeepSeek V4 FlashDeepSeekopen | 86.6% | 0731 | 59.1 | – |
| 24 | Smaug FlashUnknown | 86.2% | – | – | – |
| 25 | DeepSeek V4 ProDeepSeekopen | 85.8% | 0813 | 59.7 | – |
| 26 |  GLM-5.3Zhipu AIopen | 85.8% | glm-5.3 | – | – |
| 27 | GPT-5.6 LunaOpenAI | 85.6% | – | 56.5 | – |
| 28 | Deepseek v4 Flash VisionDeepSeek | 85.4% | – | 57.4 | – |
| 29 | Gemini 3.6 FlashGoogle | 85.2% | – | 58.6 | – |
| 30 | Claude Sonnet 4.6Anthropic | 84.8% | – | 51.8 | – |
| 31 | Gemini 3.1 ProGoogle | 84.0% | – | 60.7 | – |
| 32 | Qwen3 7Alibaba | 83.3% | – | 58.9 | – |
| 33 | GPT-5.2OpenAI | 83.2% | – | 55.5 | – |
| 34 | Kimi K2.7 CodeMoonshot AIopen | 82.8% | – | 56.2 | – |
| 35 | Smaug MiniUnknown | 82.8% | – | – | – |
| 36 | Gemini 3.5 FlashGoogle | 82.0% | – | 60.4 | – |
| 37 | GPT-5.4 nanoOpenAI | 81.1% | – | – | – |
| 38 | Claude Opus 4.5Anthropic | 80.1% | – | 56.6 | – |
| 39 | Qwen 3.8 27BAlibabaopen | 80.0% | – | 57.1 | – |
| 40 | Kimi K2.6Moonshot AIopen | 79.4% | – | – | – |
| 41 | GLM-5.2Zhipu AIopen | 78.6% | glm-5.2 | 57.4 | – |
| 42 | GPT-5.2 CodexOpenAI | 77.7% | – | 60.7 | – |
| 43 | GLM-5.3-FlashZhipu AIopen | 77.6% | – | 61.0 | – |
| 44 | Ox AlphaUnknown | 76.6% | – | – | – |
| 45 | Grok Build 0.1xAI | 76.4% | – | 56.1 | – |
| 46 | Qwen3.6 PlusAlibabaopen | 75.8% | – | 59.0 | – |
| 47 |  Nemotron 3 Ultra 550B A55BNVIDIAopen | 74.7% | – | 58.3 | – |
| 48 |  MiniMax-M3MiniMaxopen | 74.5% | minimax-m3 | 56.9 | – |
| 49 | GPT-5.4 miniOpenAI | 71.3% | – | 46.0 | – |
| 50 | Grok 4.3xAI | 70.8% | grok-4.3 | 57.8 | – |
| 51 | Qwen3.6 27BAlibabaopen | 70.3% | – | 53.2 | – |
| 52 | Gemini 3.5 Flash LiteGoogle | 60.2% | – | 44.0 | – |