LiveBench 2.0 Browser
Agents · 2026-10-14
LiveBench 2.0 Browser suite: Complete multi-step jobs in working web applications through a real browser the model sees by screenshot; 25 tasks. 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 | 100.0 |
| Opus 4.8 | 96.0 |
| Claude Fable 5 | 92.0 |
| Gemini 4 Argon | 92.0 |
| GPT-6 Sol | 92.0 |
| MiniMax M3.1 Flash | 92.0 |
| GPT-5.6 Sol | 88.0 |
| GPT-6.1 Sol | 88.0 |
| MiMo-V2.6-Pro | 88.0 |
| Claude Opus 5 | 84.0 |
| Claude Sonnet 5.5 | 84.0 |
| DeepSeek V4.1 Flash | 84.0 |
| GLM-5.3 Flash | 84.0 |
| GPT-5.6 Terra | 84.0 |
| GPT-6 Astra | 84.0 |
| Claude Haiku 5.5 | 80.0 |
| Claude Opus 5.5 | 80.0 |
| Muse Spark 1.3 | 80.0 |
| Kimi K3.1 Pro | 80.0 |
| DeepSeek V4.1 Flash 1017 | 76.0 |
| Sonnet 5 | 72.0 |
| GPT-6 Luna | 72.0 |
| Muse Wonder | 72.0 |
| Sonnet 4.6 | 68.0 |
| GPT-5.6 Luna | 68.0 |
| DeepSeek V4 Flash Vision Exp | 64.0 |
| Gemini 3.8 Flash | 64.0 |
| Grok 4.6 | 64.0 |
| Grok 4.7 | 60.0 |
| Gemini 3.7 Flash | 56.0 |
| GPT-5.5 | 48.0 |
| Qwen3.8-Flash-Next | 48.0 |
| Grok 4.5 | 44.0 |
| Haiku 4.5 | 40.0 |
| Muse Spark 1.2 | 36.0 |
| Muse Spark Open | 36.0 |
| Qwen3.8 27B | 32.0 |
| Step 5 Preview | 24.0 |
| Inkling | 16.0 |
| Kimi K3 | 16.0 |