LiveBench 2.0 Research

Open-ended research reports graded against a rubric. LiveBench 2.0, refreshed with new questions each release to limit contamination.

CategoryAgents
MetricScore (%)
SourceLiveBench (original)
LicenseNone stated
Benchmark pagelivebench.ai
StatusScored
Weight0.9 within Agents
FitDiscrimination 0.89, noise 0.39

Results

#ModelProviderResultDate
1 Claude Fable 5.2 Anthropic 94.5% 2026-11-06
2 Muse Wonder Meta 93.0% 2026-11-06
3 Gemini 4 Carbon Google 92.5% 2026-11-06
4 GPT-6 Astra OpenAI 90.6% 2026-11-06
5 GPT-6.1 Sol OpenAI 90.0% 2026-11-06
6 Grok 4.7 xAI 89.5% 2026-11-06
7 Claude Opus 5.5 Anthropic 88.9% 2026-11-06
8 Grok 4.6 xAI 88.8% 2026-11-06
9 Claude Fable 5.1 Anthropic 88.5% 2026-11-06
10 Claude Sonnet 5.5 Anthropic 88.0% 2026-11-06
11 Gemini 4 Argon Google 86.8% 2026-11-06
12 MiMo-V2.6-Pro Xiaomi 84.8% 2026-11-06
13 Claude Haiku 5.5 Anthropic 84.3% 2026-11-06
14 Muse Spark Open Meta 82.7% 2026-11-06
15 Muse Spark 1.3 Meta 82.5% 2026-11-06
16 Kimi K3.1 Pro Moonshot AI 82.1% 2026-11-06
17 Mistral Large 4 Preview Mistral 80.3% 2026-11-06
18 Kimi K3 Moonshot AI 79.8% 2026-11-06
19 GPT-5.6 Terra OpenAI 79.7% 2026-11-06
20 Gemini 3.8 Flash Google 79.4% 2026-11-06
21 GPT-6.1 Terra OpenAI 79.1% 2026-11-06
22 DeepSeek V4.1 Flash 1017 DeepSeek 78.6% 2026-11-06
23 MiniMax-M3.1-Flash MiniMax 76.7% 2026-11-06
24 Step 5 Preview StepFun 75.5% 2026-11-06
25 GLM-5.3-Flash Zhipu AI 73.9% 2026-11-06
26 Qwen3.8 Max Alibaba 72.8% 2026-11-06
27 Nemotron 3.5 Super VL 120B A12B NVIDIA 67.1% 2026-11-06
28 Qwen3.8 27B Alibaba 61.3% 2026-11-06
29 Inkling Thinking Machines 60.3% 2026-11-06
30 GPT-6 Luna OpenAI 48.5% 2026-11-06
31 Gemma 4 31B IT Google 38.1% 2026-11-06
32 Mistral Large 4 Mistral 30.8% 2026-11-06