LiveBench 2.0
Agents · 2026-10-14
LiveBench 2.0 overall: the unweighted mean of its seven suite scores (Long Agentic, Reasoning, Coding, Browser, Data Analysis, Research, Design). LiveBench 2.0 (Abacus AI, released October 14, 2026) replaced the v1 question-answer categories with seven suites, five of them agentic; each suite is one run per model at its highest supported effort, a task the run could not measure scores 0, and a model without an input a suite needs (a text-only model on Browser) scores 0 there. Not comparable with LiveBench v1.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Fable 5.1 | 88.0 |
| Claude Opus 5.5 | 86.5 |
| GPT-6 Astra | 86.1 |
| GPT-6.1 Sol | 85.4 |
| Gemini 4 Argon | 84.1 |
| Claude Fable 5 | 83.6 |
| Claude Sonnet 5.5 | 82.5 |
| Claude Opus 5 | 82.1 |
| GPT-6 Sol | 81.9 |
| DeepSeek V4.1 Flash 1017 | 79.2 |
| Muse Wonder | 78.1 |
| DeepSeek V4.1 Flash | 78.0 |
| Opus 4.8 | 76.8 |
| Kimi K3.1 Pro | 76.7 |
| GPT-5.6 Sol | 76.3 |
| Claude Haiku 5.5 | 75.6 |
| GPT-5.6 Terra | 75.0 |
| Grok 4.6 | 74.9 |
| MiMo-V2.6-Pro | 74.9 |
| Muse Spark 1.3 | 74.8 |
| Grok 4.7 | 74.1 |
| MiniMax M3.1 Flash | 73.1 |
| Sonnet 5 | 73.0 |
| GLM-5.3 Flash | 70.6 |
| Grok 4.5 | 70.1 |
| GPT-6 Luna | 69.4 |
| Qwen3.8-Flash-Next | 68.8 |
| GPT-5.5 | 68.7 |
| Gemini 3.8 Flash | 67.3 |
| Gemini 3.7 Flash | 67.2 |
| GPT-5.6 Luna | 66.1 |
| Sonnet 4.6 | 65.1 |
| DeepSeek V4 Flash Vision Exp | 63.6 |
| Kimi K3 | 63.4 |
| Qwen3.8 27B | 62.7 |
| Muse Spark 1.2 | 62.4 |
| Step 5 Preview | 62.0 |
| GLM-5.3 | 60.7 |
| DeepSeek-V4-Flash-0731 | 57.6 |
| Muse Spark Open | 57.6 |