LiveBench · Reasoning and mathematics
What it measures and how
Compute Reasoning and Mathematics category scores separately, average them with 50% weight each, then round to one decimal. This two-category average differs from selecting a single category on the official site.
How this evidence is used
The two-category average counts as one piece of ranking evidence. All specialized tasks share the LiveBench evidence family. Global Average is a cross-reference and adds no duplicate votes.
Benchmark results
Fixed rules select one representative configuration per public model. Anonymous test models are omitted. Original source ranks are retained, with up to 30 models per benchmark.
| Source rank | Source model name | Raw result | Representative configuration |
|---|---|---|---|
| 1 | gpt-6.1-sol-maxOpenAI | 94.7% | Max reasoning |
| 2 | gpt-6-astra-maxOpenAI | 94.7% | Max reasoning |
| 3 | claude-opus-5-5-max-effortAnthropic | 94.6% | Max reasoning |
| 4 | claude-fable-5-1-max-effortAnthropic | 94.3% | Max reasoning |
| 6 | gpt-5.6-sol-maxOpenAI | 93.9% | Max reasoning |
| 7 | claude-sonnet-5-5-max-effortAnthropic | 93.9% | Max reasoning |
| 9 | muse-wonder-1.0-maxMeta AI | 93.6% | Max reasoning |
| 10 | claude-opus-5-max-effortAnthropic | 93.5% | Max reasoning |
| 11 | claude-fable-5-max-effortAnthropic | 92.8% | Max reasoning |
| 12 | muse-spark-1.3-xhighMeta | 92.8% | xHigh reasoning |
| 13 | gpt-5.6-terra-maxOpenAI | 92.8% | Max reasoning |
| 14 | gpt-5.5-xhighOpenAI | 92.8% | xHigh reasoning |
| 15 | gpt-6-sol-maxOpenAI | 92.5% | Max reasoning |
| 16 | claude-opus-4-8-max-effortAnthropic | 91.8% | Max reasoning |
| 18 | grok-4.6xAI | 91.5% | Source default configuration |
| 19 | gpt-5.4-xhighOpenAI | 91.1% | xHigh reasoning |
| 20 | claude-sonnet-5-xhigh-effortAnthropic | 90.8% | xHigh reasoning |
| 21 | gemini-3.7-flash-highGoogle | 90.6% | High reasoning |
| 22 | muse-spark-1.2-xhighMeta | 90.6% | xHigh reasoning |
| 23 | deepseek-v4-pro-0813DeepSeek | 90.5% | Source default configuration |
| 24 | gemini-3.8-flash-highGoogle | 90.4% | High reasoning |
| 25 | claude-opus-4-7-xhigh-effortAnthropic | 90.0% | xHigh reasoning |
| 26 | deepseek-v4.1-flash-maxDeepSeek | 90.0% | Source default configuration |
| 27 | qwen3.8-maxAlibaba | 89.8% | Source default configuration |
| 28 | grok-4.7-xhighxAI | 89.2% | xHigh reasoning |
| 29 | grok-4.5xAI | 89.0% | Source default configuration |
| 30 | claude-opus-4-6-thinking-auto-high-effortAnthropic | 89.0% | High reasoning |
| 31 | mistral-large-4-high— | 88.8% | Source default configuration |
| 32 | gpt-5.2-2025-12-11-highOpenAI | 88.2% | High reasoning |
| 33 | smaug-flashAbacus.AI | 88.1% | Source default configuration |
Limitations and attribution
The two-category average measures combined performance, not Reasoning alone. Switching categories or including fine-tuned models on the official page changes scores and ranks. Participating organization Abacus.AI also develops Smaug; results exclusive to its models need support from other independent sources.
Data license: Apache 2.0
LiveBench publishes the component scores. e/acc averages the two categories above with equal weight and ranks the results. The two-category average and the e/acc consensus index use different scales and cannot be added directly.