New: Agent Index for knowledge work is live.Explore the index ↗

Best LLMs for research reports

LiveBench: Research

LiveBench 2.0 evaluates research reports on its October 25, 2026 release. This suite contains 22 tasks or questions, scored from 0 to 100 in one run per model. Models use their highest supported effort unless a different setting is reported. This is an individual task-suite score, separate from LiveBench’s weighted overall average.

Updated: Nov 5, 2026Data as of: Nov 2, 2026

Widely usedNot saturated

Official benchmark

LiveBench: Research results. Each row identifies its benchmark version and evaluation track. Scores from different versions are not directly comparable. Best effort from the latest stored report date per model and version.
Model / creator	Version	Reported score
Muse Wonder 1.0Meta · Proprietary	2.0 / 2026-10-25	88%
Grok 4.7xAI · Proprietary	2.0 / 2026-10-25	86.4%
GPT-6 AstraOpenAI · Proprietary	2.0 / 2026-10-25	85.2%
Claude Opus 5.5Anthropic · Proprietary	2.0 / 2026-10-25	85%
Claude Sonnet 5.5Anthropic · Proprietary	2.0 / 2026-10-25	83.6%
GPT-6.1 SolOpenAI · Proprietary	2.0 / 2026-10-25	83.5%
Gemini 4 ArgonGoogle · Proprietary	2.0 / 2026-10-25	83.1%
Claude Fable 5.1Anthropic · Proprietary	2.0 / 2026-10-25	82%
Claude Haiku 5.5Anthropic · Proprietary	2.0 / 2026-10-25	81.9%
Muse Spark OpenMeta · Open source / open weights	2.0 / 2026-10-25	80.3%
DeepSeek V4.1 Flash 1017DeepSeek · Open source / open weights	2.0 / 2026-10-25	79.5%
Gemini 3.8 FlashGoogle · Proprietary	2.0 / 2026-10-25	79.1%
Kimi K3Moonshot AI · Open source / open weights	2.0 / 2026-10-25	79%
Kimi K3.1 ProMoonshot AI · Open source / open weights	2.0 / 2026-10-25	77.9%
MiMo-V2.6-ProXiaomi · Open source / open weights	2.0 / 2026-10-25	76.7%
GPT-6.1 TerraOpenAI · Proprietary	2.0 / 2026-10-25	73.9%
Step 5 PreviewStepFun · Open source / open weights	2.0 / 2026-10-25	72.8%
MiniMax M3.1 FlashMiniMax · Open source / open weights	2.0 / 2026-10-25	72.7%
GLM-5.3-FlashZ.ai · Open source / open weights	2.0 / 2026-10-25	70.1%
Muse Spark 1.3Meta · Proprietary	2.0 / 2026-10-25	68.6%
Nemotron 3.5 Super VLNVIDIA · Open source / open weights	2.0 / 2026-10-25	66.7%
GPT-5.6 TerraOpenAI · Proprietary	2.0 / 2026-10-25	60.4%
InklingThinking Machines · Open source / open weights	2.0 / 2026-10-25	59.9%
Qwen3.8 27BQwen · Open source / open weights	2.0 / 2026-10-25	59.4%
GPT-6 LunaOpenAI · Proprietary	2.0 / 2026-10-25	43.5%

View all 25 published reports

Muse Wonder 1.088%

2.0 / 2026-10-25

release: 2026-10-25 · Highest supported effort unless publisher reports an exception; single run · suite: research

Claude Haiku 5.581.9%

2.0 / 2026-10-25

release: 2026-10-25 · Highest supported effort unless publisher reports an exception; single run · suite: research

Muse Spark Open80.3%

2.0 / 2026-10-25

release: 2026-10-25 · Highest supported effort unless publisher reports an exception; single run · suite: research

MiMo-V2.6-Pro76.7%

2.0 / 2026-10-25

release: 2026-10-25 · Highest supported effort unless publisher reports an exception; single run · suite: research

GPT-6.1 Terra73.9%

2.0 / 2026-10-25

release: 2026-10-25 · Highest supported effort unless publisher reports an exception; single run · suite: research

Step 5 Preview72.8%

2.0 / 2026-10-25

release: 2026-10-25 · Highest supported effort unless publisher reports an exception; single run · suite: research

Nemotron 3.5 Super VL66.7%

2.0 / 2026-10-25

release: 2026-10-25 · Highest supported effort unless publisher reports an exception; single run · suite: research

GPT-5.6 Terra60.4%

2.0 / 2026-10-25

release: 2026-10-25 · Highest supported effort unless publisher reports an exception; single run · suite: research

2.0 / 2026-10-25

release: 2026-10-25 · Highest supported effort unless publisher reports an exception; single run · suite: research

Qwen3.8 27B59.4%

2.0 / 2026-10-25

release: 2026-10-25 · Highest supported effort unless publisher reports an exception; single run · suite: research

GPT-6 Luna43.5%

2.0 / 2026-10-25

release: 2026-10-25 · Highest supported effort unless publisher reports an exception; single run · suite: research

2.0 / 2026-10-25

release: 2026-10-25 · Highest supported effort unless publisher reports an exception; single run · suite: research

GPT-6 Astra85.2%

2.0 / 2026-10-25

release: 2026-10-25 · Highest supported effort unless publisher reports an exception; single run · suite: research

Claude Opus 5.585%

2.0 / 2026-10-25

release: 2026-10-25 · Highest supported effort unless publisher reports an exception; single run · suite: research

Claude Sonnet 5.583.6%

2.0 / 2026-10-25

release: 2026-10-25 · Highest supported effort unless publisher reports an exception; single run · suite: research

GPT-6.1 Sol83.5%

2.0 / 2026-10-25

release: 2026-10-25 · Highest supported effort unless publisher reports an exception; single run · suite: research

Gemini 4 Argon83.1%

2.0 / 2026-10-25

release: 2026-10-25 · Highest supported effort unless publisher reports an exception; single run · suite: research

Claude Fable 5.182%

2.0 / 2026-10-25

release: 2026-10-25 · Highest supported effort unless publisher reports an exception; single run · suite: research

DeepSeek V4.1 Flash 101779.5%

2.0 / 2026-10-25

release: 2026-10-25 · Highest supported effort unless publisher reports an exception; single run · suite: research

Gemini 3.8 Flash79.1%

2.0 / 2026-10-25

release: 2026-10-25 · Highest supported effort unless publisher reports an exception; single run · suite: research

2.0 / 2026-10-25

release: 2026-10-25 · Highest supported effort unless publisher reports an exception; single run · suite: research

Kimi K3.1 Pro77.9%

2.0 / 2026-10-25

release: 2026-10-25 · Highest supported effort unless publisher reports an exception; single run · suite: research

MiniMax M3.1 Flash72.7%

2.0 / 2026-10-25

release: 2026-10-25 · Highest supported effort unless publisher reports an exception; single run · suite: research

GLM-5.3-Flash70.1%

2.0 / 2026-10-25

release: 2026-10-25 · Highest supported effort unless publisher reports an exception; single run · suite: research

Muse Spark 1.368.6%

2.0 / 2026-10-25

release: 2026-10-25 · Highest supported effort unless publisher reports an exception; single run · suite: research