All tests
Language
LiveBench language
Language puzzles, connections and text understanding.
Published by LiveBench under Apache-2.0. Compare it with other tests
Top score90.7%Claude Fable 5 (max effort), Anthropic
Models tested63from 11 makers
Median score79.7%Half the models score above it
Scale0 to 100%The share of tasks done right
Top 15
Each model at its best configuration, coloured by maker.
The best model from each maker
Each maker’s highest-scoring model on this test, and its rank.
| Maker | Best model | Score | Rank |
|---|---|---|---|
 Anthropic | Claude Fable 5 (max effort) | 90.7% | #1 of 63 |
 OpenAI | GPT-6.1 Sol (max) | 90.1% | #2 of 63 |
 Google | Gemini 3.8 Flash (high) | 87.8% | #6 of 63 |
 Moonshot AI | Kimi K3 | 85.5% | #10 of 63 |
 xAI | Grok 4.6 | 83.7% | #17 of 63 |
 DeepSeek | DeepSeek V4 Pro 0813 | 82.1% | #24 of 63 |
 Z.ai | GLM-5.3 | 79.9% | #29 of 63 |
 Alibaba | Qwen3.7 Max | 79.7% | #31 of 63 |
 MiniMax | MiniMax-M3 | 76.8% | #42 of 63 |
 NVIDIA | Nemotron 3 Ultra 550B A55B | 70.8% | #59 of 63 |
Every result
63 models, each once at its best configuration, highest score first. Models with the same score share a rank.
| Rank | Model | Maker | Score |
|---|---|---|---|
| #1 | Claude Fable 5 (max effort) | Anthropic | 90.7% |
| #2 | GPT-6.1 Sol (max) | OpenAI | 90.1% |
| #3 | Claude Fable 5.1 (max effort) | Anthropic | 89.5% |
| #4 | GPT-6 Astra (max) | OpenAI | 89.4% |
| #5 | Claude Opus 5 (max effort) | Anthropic | 88.7% |
| #6 | Gemini 3.8 Flash (high) | Google | 87.8% |
| #7 | GPT-5.6 Sol (max) | OpenAI | 87.7% |
| #8 | GPT-5.5 (xhigh) | OpenAI | 87.4% |
| #9 | Claude Opus 5.5 (max effort) | Anthropic | 86.3% |
| #10 | Kimi K3 | Moonshot AI | 85.5% |
| #11 | Gemini 3.7 Flash (high) | Google | 85.5% |
| #12 | Gemini 3.1 Pro | Google | 85.4% |
| #13 | GPT-6 Sol (max) | OpenAI | 85.3% |
| #14 | Gemini 3.5 Flash (high) | Google | 84.6% |
| #15 | Smaug Agentic |  Not reported | 84.4% |
| #16 | Gemini 3.6 Flash (high) | Google | 83.9% |
| #17 | Grok 4.6 | xAI | 83.7% |
| #18 | Claude Sonnet 5.5 (xhigh effort) | Anthropic | 83.4% |
| #19 | Claude Opus 4.6 (thinking auto high effort) | Anthropic | 83.3% |
| #20 | GPT-5.6 Terra (max) | OpenAI | 82.9% |
| #21 | Grok 4.5 | xAI | 82.8% |
| #22 | Muse Spark 1.3 (xhigh) | Meta AI | 82.8% |
| #23 | GPT-5.4 (xhigh) | OpenAI | 82.6% |
| #24 | DeepSeek V4 Pro 0813 | DeepSeek | 82.1% |
| #25 | Claude Opus 4.5 (20251101) (thinking 64k high effort) | Anthropic | 81.3% |
How it is scored
Arithmetic mean of the 3 published task scores in this category. Only complete results are ranked. Exact model and effort configurations are preserved.
Results from LiveBench (Apache-2.0), as published. Model names link to their pages where the site can match them; makers come from the model’s page, or else from the publisher.