BenchLeader
Menu
Best LLMsChoose a modelLeaderboardsModelsBenchmarksProvidersReleases15

LiveBench Research

LiveBench 2.0 research suite.

As of 21 Oct 2026, Claude Opus 5 leads LiveBench Research on BenchLeader with 88.9%, ahead of Muse Wonder at 88.0%, across 48 model configurations with a published result.

Published by
LiveBench
Category
Knowledge
Index weight
Reference only
Models
48
Data as of
21 Oct 2026

LiveBench (livebench.ai), White et al. 2024.

What the test looks like

LiveBench's research suite: assembling and checking findings across sources into a graded report.

How it is scored

Mean task score for the suite, 0–100, from a single run.

What to keep in mind

Judges both coverage and correctness; citation style and formatting can cost points.

Top 15
  1. 1Claude Opus 588.9%
  2. 2Muse Wonder88.0%
  3. 3Claude Fable 586.9%
  4. 4DeepSeek V4.1 Flash86.8%
  5. 5Grok 4.786.4%
  6. 6Grok 4.585.9%
  7. 7GPT-6 Astra85.2%
  8. 8Claude Opus 5.585.0%
  9. 9Grok 4.684.5%
  10. 10Claude Sonnet 5.583.6%
  11. 11GPT-6.1 Sol83.5%
  12. 12Gemini 4 Argon83.1%
  13. 13Claude Sonnet 582.8%
  14. 14Claude Fable 5.182.0%
  15. 15Claude Haiku 5.581.9%
48 of 48
#
1Claude Opus 588.9%65.42026-10-14
2Muse Wonder88.0%63.62026-10-14
3Claude Fable 586.9%67.12026-10-14
4DeepSeek V4.1 Flash86.8%–2026-10-14
5Grok 4.786.4%–2026-10-14
6Grok 4.585.9%58.92026-10-14
7GPT-6 Astra85.2%67.92026-10-14
8Claude Opus 5.585.0%61.92026-10-14
9Grok 4.684.5%57.82026-10-14
10Claude Sonnet 5.583.6%–2026-10-14
11GPT-6.1 Sol83.5%63.32026-10-14
12Gemini 4 Argon83.1%63.32026-10-14
13Claude Sonnet 582.8%56.02026-10-14
14Claude Fable 5.182.0%66.22026-10-14
15Claude Haiku 5.581.9%–2026-10-14
16Claude Opus 4.881.8%60.12026-10-14
17NewMuse Spark Open80.3%–2026-10-14
18NewDeepSeek V4.1 Flash 101779.5%–2026-10-14
19Gemini 3.8 Flash79.1%55.62026-10-14
20Kimi K379.0%59.12026-10-14
21MiMo-V2.6-Pro76.7%64.12026-10-14
22Gemini 3.7 Flash76.5%53.22026-10-14
23Qwen3.8 Flash74.6%–2026-10-14
24GLM 5.373.2%–2026-10-14
25Step 5 Preview72.8%60.62026-10-14
26NewBeam 1.072.8%54.62026-10-14
27NewMiniMax-M3.1-Flash72.7%61.92026-10-14
28Nemotron Super VL 3.570.9%–2026-10-14
29GLM 5.3 Flash70.1%62.72026-10-14
30Qwen3.8 Maxmax69.9%63.32026-10-14
31GPT-6 Sol69.3%–2026-10-14
32Deepseek v4 Flash Vision69.2%–2026-10-14
33Muse Spark 1.368.6%58.32026-10-14
34MiniMax M368.0%56.52026-10-14
35Muse Spark 1.267.7%–2026-10-14
36Gemini 3.1 Pro65.7%60.72026-10-14
37Claude Sonnet 4.663.6%57.12026-10-14
38GPT-5.561.6%60.22026-10-14
39GPT-5.6 Terra60.4%52.62026-10-14
40Qwen3.8 27B59.4%54.52026-10-14
41GPT-5.6 Sol54.5%55.92026-10-14
42GPT-5.6 Luna51.0%47.92026-10-14
43Gemma 4 31B44.4%48.32026-10-14
44GPT-6 Luna43.5%–2026-10-14
45Nemotron Ultra 343.2%–2026-10-14
46Claude Haiku 4.542.2%45.32026-10-14
47Nemotron Lightning 3.540.0%–2026-10-14
48Mistral Large 428.1%56.72026-10-14

More knowledge benchmarks · All LLM benchmarks

Cite as: BenchLeader, “LiveBench Research leaderboard”, https://www.benchleader.com/benchmarks/livebench_research, data as of 21 Oct 2026.

LiveBench Research: questions

What does LiveBench Research measure?
LiveBench's research suite: assembling and checking findings across sources into a graded report. Scores are reported in percent of tasks solved; higher is better.
Which AI model leads LiveBench Research?
Claude Opus 5 leads LiveBench Research with 88.9% as of 21 Oct 2026, ahead of Muse Wonder at 88.0%.
How many models have LiveBench Research results?
48 model configurations have a LiveBench Research result on BenchLeader, all taken from LiveBench.
Who runs LiveBench Research and how often is it updated?
LiveBench Research is published by LiveBench. BenchLeader re-reads the published results once a day and records the date each result was published.
Does LiveBench Research count toward the BenchLeader Index?
No. LiveBench Research is shown for reference but left out of the composite index.