All tests
Language
LiveBench instructions
Following constraints in rewriting, summaries and generation.
Published by LiveBench under Apache-2.0. Compare it with other tests
Top score81.4%Gemini 3.8 Flash (high), Google
Models tested63from 11 makers
Median score69.3%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 |
|---|---|---|---|
 Google | Gemini 3.8 Flash (high) | 81.4% | #1 of 63 |
 Alibaba | Qwen3.8 Flash Next | 77.1% | #5 of 63 |
 Anthropic | Claude Fable 5 (max effort) | 75.8% | #6 of 63 |
 OpenAI | GPT-6 Astra (max) | 75.6% | #8 of 63 |
 xAI | Grok 4.7 (xhigh) | 75.3% | #10 of 63 |
 NVIDIA | Nemotron 3 Ultra 550B A55B | 73.5% | #16 of 63 |
 Moonshot AI | Kimi K3 | 71.4% | #23 of 63 |
 DeepSeek | DeepSeek V4 Flash Vision Exp | 71.0% | #25 of 63 |
 Z.ai | GLM-5.3 | 69.3% | #32 of 63 |
 MiniMax | MiniMax-M3 | 57.5% | #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 | Gemini 3.8 Flash (high) | Google | 81.4% |
| #2 | Gemini 3.7 Flash (high) | Google | 79.9% |
| #3 | Gemini 3.1 Pro | Google | 79.1% |
| #4 | Muse Spark 1.3 (xhigh) |  Meta AI | 78.0% |
| #5 | Qwen3.8 Flash Next | Alibaba | 77.1% |
| #6 | Claude Fable 5 (max effort) | Anthropic | 75.8% |
| #7 | Gemini 3.5 Flash (high) | Google | 75.6% |
| #8 | GPT-6 Astra (max) | OpenAI | 75.6% |
| #9 | Gemini 3.6 Flash (high) | Google | 75.4% |
| #10 | Grok 4.7 (xhigh) | xAI | 75.3% |
| #11 | Muse Spark 1.2 (xhigh) | Meta AI | 74.3% |
| #12 | GPT-6.1 Sol (max) | OpenAI | 74.2% |
| #13 | Qwen3.8 Max | Alibaba | 74.1% |
| #14 | Qwen3.7 Max | Alibaba | 74.0% |
| #15 | Smaug Mini | Not reported | 73.9% |
| #16 | Nemotron 3 Ultra 550B A55B | NVIDIA | 73.5% |
| #17 | Claude Fable 5.1 (max effort) | Anthropic | 73.0% |
| #18 | Qwen3.8 27B | Alibaba | 72.7% |
| #19 | Claude Opus 4.8 (max effort) | Anthropic | 72.0% |
| #20 | Grok 4.6 | xAI | 71.9% |
| #21 | GPT-5.6 Sol (max) | OpenAI | 71.8% |
| #22 | Grok 4.5 | xAI | 71.5% |
| #23 | Kimi K3 | Moonshot AI | 71.4% |
| #24 | Smaug Agentic | Not reported | 71.0% |
| #25 | DeepSeek V4 Flash Vision Exp | DeepSeek | 71.0% |
How it is scored
Arithmetic mean of the 4 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.