BenchLeader
Menu
Best LLMsChoose a modelLeaderboardsModelsBenchmarksProvidersReleases9

LiveBench 1.0 Reasoning

Reasoning category of LiveBench 1.0, frozen at its final release (25 June 2026). LiveBench 2.0's Reasoning suite is a different test.

As of 26 Oct 2026, GPT-6 Astra leads LiveBench 1.0 Reasoning on BenchLeader with 92.7%, ahead of GPT-6.1 Sol at 92.6%, across 67 model configurations with a published result.

Published by
LiveBench
Category
Reasoning
Index weight
Reference only
Models
67
Data as of
26 Oct 2026
Newest result
25 Jun 2026

The publisher has posted no new results on this board since 25 Jun 2026. BenchLeader re-reads it every day, so newer models appear here as soon as they are published.

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

This board is frozen. A different, newer test now carries its name, and its scores are on LiveBench 2.0 Reasoning.

What the test looks like

LiveBench 1.0's reasoning questions: logic puzzles, spatial problems and variants of classic riddles, freshly written so the answers are not memorised.

How it is scored

Percent correct, graded automatically, from LiveBench 1.0's final release (25 June 2026). LiveBench 2.0's Reasoning suite is a different test, so these scores no longer update.

What to keep in mind

A small category, so a few questions move it, and the puzzle styles are specific enough that models can be tuned to them.

Top 15
  1. 1GPT-6 Astra (max)92.7%
  2. 2GPT-6.1 Sol (max)92.6%
  3. 3Claude Opus 5.5 (max)92.2%
  4. 4Claude Fable 5.1 (max)91.7%
  5. 5GPT-5.6 Sol (max)91.7%
  6. 6Claude Sonnet 5.5 (max)91.6%
  7. 7GPT-6.1 Sol (xhigh)91.6%
  8. 8Claude Opus 5 (max)91.2%
  9. 9Kimi K390.7%
  10. 10Claude Opus 5.5 (xhigh)90.7%
  11. 11Muse Wonder (max)90.6%
  12. 12GPT-5.6 Terra (max)90.6%
  13. 13Grok 4.690.5%
  14. 14Muse Spark 1.2 (xhigh)90.0%
  15. 15Claude Fable 5 (max)89.7%
67 of 67
#
1GPT-6 Astramax92.7%70.42026-06-25
2GPT-6.1 Solmax92.6%69.92026-06-25
3Claude Opus 5.5max92.2%72.62026-06-25
4Claude Fable 5.1max91.7%70.52026-06-25
5GPT-5.6 Solmax91.7%68.02026-06-25
6Claude Sonnet 5.5max91.6%69.42026-06-25
7GPT-6.1 Solxhigh91.6%67.72026-06-25
8Claude Opus 5max91.2%68.82026-06-25
9Kimi K390.7%59.22026-06-25
10Claude Opus 5.5xhigh90.7%70.02026-06-25
11Muse Wondermax90.6%67.52026-06-25
12GPT-5.6 Terramax90.6%64.42026-06-25
13Grok 4.690.5%57.12026-06-25
14Muse Spark 1.2xhigh90.0%63.22026-06-25
15Claude Fable 5max89.7%69.32026-06-25
16GPT-5.5xhigh89.7%66.12026-06-25
17Muse Spark 1.3xhigh89.7%64.92026-06-25
18Gemini 3.8 Flashhigh89.3%64.42026-06-25
19Claude Opus 4.8max89.2%63.32026-06-25
20Claude Sonnet 5xhigh88.7%59.02026-06-25
21Claude Opus 4.6high88.7%62.72026-06-25
22GPT-6 Solmax88.7%66.42026-06-25
23Qwen3.8 Maxmax88.2%63.22026-06-25
24GPT-5.4xhigh88.1%64.72026-06-25
25Gemini 3.7 Flashhigh87.8%63.52026-06-25
26Muse Spark 1.1xhigh87.7%61.62026-06-25
27Qwen3.8-Flash-Next87.4%60.32026-06-25
28Claude Opus 4.7xhigh87.2%57.42026-06-25
29Grok 4.587.2%60.02026-06-25
30Claude Sonnet 5.5xhigh86.8%67.92026-06-25
31DeepSeek V4.1 Flashmax86.7%61.22026-06-25
32DeepSeek V4 Flash86.6%54.32026-06-25
33Smaug Flash86.2%–2026-06-25
34DeepSeek V4 Pro85.8%53.42026-06-25
35GLM 5.385.8%–2026-06-25
36GPT-5.6 Lunamax85.6%59.92026-06-25
37Deepseek v4 Flash Vision85.4%–2026-06-25
38Gemini 3.6 Flashhigh85.2%62.02026-06-25
39Claude Sonnet 4.6medium84.8%–2026-06-25
40Claude Haiku 5.5max84.6%60.52026-06-25
41Gemini 3.1 Prohigh84.0%60.32026-06-25
42Mistral Large 4high83.9%49.82026-06-25
43Qwen3.7max83.3%60.62026-06-25
44GPT-5.2high83.2%56.02026-06-25
45Kimi K2.7 Code82.8%55.82026-06-25
46Smaug Mini82.8%–2026-06-25
47Grok 4.7xhigh82.7%61.62026-06-25
48Gemini 3.5 Flashhigh82.0%63.12026-06-25
49GPT-6 Lunamax81.8%60.02026-06-25
50Claude Haiku 5.5xhigh81.2%61.52026-06-25
51GPT-5.4 Nanoxhigh81.1%54.22026-06-25
52Claude Opus 4.5high80.1%59.22026-06-25
53Qwen3.8 27B80.0%54.32026-06-25
54Kimi K2.6thinking79.4%59.42026-06-25
55GLM 5.278.6%51.82026-06-25
56GPT-5.2-Codex77.7%57.32026-06-25
57GLM 5.3 Flash77.6%62.32026-06-25
58Ox Alphamax76.6%–2026-06-25
59Grok Build 0.176.4%54.82026-06-25
60Qwen3.6 Plus75.8%58.82026-06-25

More reasoning benchmarks · All LLM benchmarks

Cite as: BenchLeader, “LiveBench 1.0 Reasoning leaderboard”, data as of 26 Oct 2026, https://www.benchleader.com/benchmarks/livebench_reasoning.

LiveBench 1.0 Reasoning: questions

What does LiveBench 1.0 Reasoning measure?
LiveBench 1.0's reasoning questions: logic puzzles, spatial problems and variants of classic riddles, freshly written so the answers are not memorised. Scores are reported in percent of tasks solved; higher is better.
Which AI model leads LiveBench 1.0 Reasoning?
GPT-6 Astra leads LiveBench 1.0 Reasoning with 92.7% as of 26 Oct 2026, ahead of GPT-6.1 Sol at 92.6%.
How many models have LiveBench 1.0 Reasoning results?
67 model configurations have a LiveBench 1.0 Reasoning result on BenchLeader, all taken from LiveBench.
Who runs LiveBench 1.0 Reasoning and how often is it updated?
LiveBench 1.0 Reasoning is published by LiveBench. BenchLeader re-reads the published results once a day and records the date each result was published.
Does LiveBench 1.0 Reasoning count toward the BenchLeader Index?
No. LiveBench 1.0 Reasoning is shown for reference but left out of the composite index.