BenchLeader
Menu
Best LLMsChoose a modelLeaderboardsModelsBenchmarksProvidersReleases12

LiveCodeBench

Competitive programming problems released after model cutoffs. Run by Vals AI.

As of 24 Oct 2026, Claude Fable 5.1 leads LiveCodeBench on BenchLeader with 90.5%, ahead of Claude Fable 5 at 89.8%, across 137 model configurations with a published result.

Published by
Vals AI
Category
Coding
Index weight
1.0
Models
137
Data as of
24 Oct 2026
Newest result
1 Sept 2026

Vals AI (vals.ai).

What the test looks like

LiveCodeBench collects competitive-programming problems published after each model's training cutoff and checks solutions by running hidden tests.

How it is scored

Percent of problems passing all tests; Vals AI runs a fixed recent window.

What to keep in mind

Only the window after a model's cutoff is contamination-free, and the window changes over time, so scores across dates are not strictly comparable.

Top 15
  1. 1Claude Fable 5.190.5%
  2. 2Claude Fable 589.8%
  3. 3Gemini 3.8 Flash (high)89.5%
  4. 4Claude Opus 589.0%
  5. 5Gemini 3.7 Flash (high)88.7%
  6. 6Gemini 3.1 Pro (high)88.5%
  7. 7Grok 4.6 (high)88.2%
  8. 8Gemini 3.6 Flash (high)88.1%
  9. 9GPT-5.2-Codex88.0%
  10. 10Qwen3.8 Max (max)87.8%
  11. 11Claude Opus 4.887.8%
  12. 12Gemini 3.5 Flash (high)87.6%
  13. 13DeepSeek V4 Pro (max)87.5%
  14. 14Grok 4.5 (high)87.3%
  15. 15GPT-5.3-Codex (xhigh)87.3%
137 of 137
#
1Claude Fable 5.190.5%66.42026-09-01
2Claude Fable 589.8%67.12026-09-01
3Gemini 3.8 Flashhigh89.5%64.42026-09-01
4Claude Opus 589.0%65.52026-09-01
5Gemini 3.7 Flashhigh88.7%63.52026-09-01
6Gemini 3.1 Prohigh88.5%60.32026-09-01
7Grok 4.6high88.2%62.92026-09-01
8Gemini 3.6 Flashhigh88.1%62.02026-09-01
9GPT-5.2-Codex88.0%57.32026-09-01
10Qwen3.8 Maxmax87.8%63.32026-09-01
11Claude Opus 4.887.8%60.12026-09-01
12Gemini 3.5 Flashhigh87.6%63.12026-09-01
13DeepSeek V4 Promax87.5%60.82026-09-01
14Grok 4.5high87.3%59.32026-09-01
15GPT-5.3-Codexxhigh87.3%63.82026-09-01
16DeepSeek V4 Flashhigh87.3%57.42026-09-01
17Kimi K3max87.2%64.72026-09-01
18Qwen3.7max87.1%60.62026-09-01
19Kimi K2.686.8%55.42026-09-01
20GPT-5 minihigh86.6%54.12026-09-01
21GPT-5.1high86.5%57.72026-09-01
22Gemini 3 Pro86.4%60.42026-09-01
23Nemotron 3 Ultra 550B A55B86.0%55.52026-09-01
24Qwen3.6 Plus86.0%58.82026-09-01
25GPT-5.6 Terraxhigh85.9%62.82026-09-01
26Inkling Small85.9%54.82026-09-01
27Muse Spark 1.1xhigh85.9%61.62026-09-01
28GPT-5high85.9%57.92026-09-01
29Gemini 3 Flashhigh85.6%58.32026-09-01
30GPT-5.1-Codex85.5%–2026-09-01
31GPT-5.2xhigh85.4%60.72026-09-01
32Qwen3.5 Plusthinking85.3%57.32026-09-01
33GPT-5.5xhigh85.3%66.12026-09-01
34Claude Opus 4.785.1%63.32026-09-01
35GPT-5-Codexhigh84.7%57.52026-09-01
36Claude Opus 4.6thinking84.7%62.52026-09-01
37Grok 4.3high84.5%57.02026-09-01
38Grok 4.20thinking84.3%58.62026-09-01
39GPT-5.4xhigh84.1%64.72026-09-01
40GPT-5.4 Nanohigh84.0%48.42026-09-01
41Qwen3.8 27Bxhigh84.0%56.22026-09-01
42Ling 3.0 Flash84.0%52.82026-09-01
43o3high83.9%53.92026-09-01
44Kimi K2.5thinking83.9%57.62026-09-01
45Claude Opus 4.5thinking83.7%59.92026-09-01
46GPT-5.1-Codex-Maxmax83.6%–2026-09-01
47Qwen3.5 Flash83.3%52.02026-09-01
48Grok 483.3%58.02026-09-01
49gpt-oss-120b83.2%45.62026-09-01
50GPT-5.6 Solmax82.6%68.02026-09-01
51Claude Sonnet 582.4%56.12026-09-01
52GLM 4.782.2%51.92026-09-01
53o4-minihigh82.2%50.12026-09-01
54MiniMax M382.2%56.42026-09-01
55Claude Sonnet 4.682.1%57.12026-09-01
56Kimi K2.7 Code82.0%55.82026-09-01
57GLM-5thinking81.9%58.92026-09-01
58MiniMax-M2.181.8%53.82026-09-01
59GPT-5.4 Minixhigh81.5%56.32026-09-01
60GLM 5.181.4%55.92026-09-01

More coding benchmarks · All LLM benchmarks

Cite as: BenchLeader, “LiveCodeBench leaderboard”, data as of 24 Oct 2026, https://www.benchleader.com/benchmarks/vals_livecodebench.

LiveCodeBench: questions

What does LiveCodeBench measure?
LiveCodeBench collects competitive-programming problems published after each model's training cutoff and checks solutions by running hidden tests. Scores are reported in percent of tasks solved; higher is better.
Which AI model leads LiveCodeBench?
Claude Fable 5.1 leads LiveCodeBench with 90.5% as of 24 Oct 2026, ahead of Claude Fable 5 at 89.8%.
How many models have LiveCodeBench results?
137 model configurations have a LiveCodeBench result on BenchLeader, all taken from Vals AI.
Who runs LiveCodeBench and how often is it updated?
LiveCodeBench is published by Vals AI. BenchLeader re-reads the published results once a day and records the date each result was published.
Does LiveCodeBench count toward the BenchLeader Index?
Yes. LiveCodeBench contributes to the coding category of the BenchLeader Index, normalised so that 50 is the average of the evaluated models.