LiveBench Global Average
What it measures and how
Keep the official Global Average as an overall cross-reference. Ranking uses nonoverlapping specialized tasks.
How this evidence is used
The raw Global Average remains a cross-reference. Ranking votes come from nonoverlapping specialized tasks.
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 | claude-fable-5-1-max-effortAnthropic | 83.4% | Max reasoning |
| 2 | claude-opus-5-5-max-effortAnthropic | 83.2% | Max reasoning |
| 3 | claude-fable-5-max-effortAnthropic | 83.0% | Max reasoning |
| 4 | gpt-6-astra-maxOpenAI | 82.2% | Max reasoning |
| 6 | gpt-6.1-sol-maxOpenAI | 81.6% | Max reasoning |
| 7 | muse-spark-1.3-xhighMeta | 81.6% | xHigh reasoning |
| 8 | muse-wonder-1.0-maxMeta AI | 81.2% | Max reasoning |
| 10 | deepseek-v4.1-flash-maxDeepSeek | 81.1% | Source default configuration |
| 11 | gpt-5.6-sol-maxOpenAI | 81.1% | Max reasoning |
| 12 | gpt-5.5-xhighOpenAI | 80.2% | xHigh reasoning |
| 13 | claude-opus-5-max-effortAnthropic | 80.1% | Max reasoning |
| 14 | smaug-agenticAbacus.AI | 79.5% | Source default configuration |
| 15 | gpt-6-sol-maxOpenAI | 79.2% | Max reasoning |
| 16 | kimi-k3Moonshot AI | 79.2% | Source default configuration |
| 17 | gemini-3.7-flash-highGoogle | 78.8% | High reasoning |
| 18 | qwen3.8-maxAlibaba | 78.5% | Source default configuration |
| 19 | grok-4.6xAI | 78.0% | Source default configuration |
| 20 | gpt-5.4-xhighOpenAI | 78.0% | xHigh reasoning |
| 21 | muse-spark-1.2-xhighMeta | 78.0% | xHigh reasoning |
| 22 | gpt-5.6-terra-maxOpenAI | 77.9% | Max reasoning |
| 24 | deepseek-v4-pro-0813DeepSeek | 77.4% | Source default configuration |
| 25 | smaug-flashAbacus.AI | 77.4% | Source default configuration |
| 26 | grok-4.7-xhighxAI | 77.4% | xHigh reasoning |
| 27 | gemini-3.1-pro-preview-highGoogle | 77.0% | High reasoning |
| 28 | smaug-miniAbacus.AI | 76.9% | Source default configuration |
| 29 | deepseek-v4-flash-vision-expDeepSeek | 76.8% | Source default configuration |
| 30 | claude-opus-4-7-xhigh-effortAnthropic | 76.5% | xHigh reasoning |
| 31 | claude-opus-4-8-max-effortAnthropic | 76.2% | Max reasoning |
| 32 | qwen3.8-flash-nextAlibaba | 76.2% | Source default configuration |
| 33 | glm-5.3Z.ai | 76.1% | Source default configuration |
Limitations and attribution
An overall average can dilute changes in specific capabilities. It is a cross-reference and adds no duplicate votes. Participating organization Abacus.AI also develops Smaug; results exclusive to its models need support from other independent sources.
Data license: Apache 2.0
Results are published by LiveBench. Raw scores and the e/acc consensus index use different scales and cannot be added directly.