LiveCodeBench
Contamination-limited competitive programming problems
codingactive
Score in %
Definition
- Version
- not in the notes
- Category
- coding
- Maintainer
- UC Berkeley/MIT/Cornell authors, USA
- Status
- active
- Home page
- https://arxiv.org/abs/2403.07974
Sources
- Claude Fable 5.1 Benchmarks, Cost and Capabilities | Vals AI vals.ai https://vals.ai/models/anthropic_claude-fable-5-1
- DeepSeek V4-Pro Review: Benchmarks, Pricing & Architecture blog.buildfastwithai.com https://blog.buildfastwithai.com/deepseek-v4-pro-review-2026
- Qwen/Qwen3.8-27B , Hugging Face huggingface.co https://huggingface.co/Qwen/Qwen3.8-27B
- huggingface.co/JetBrains/Mellum2.1-12B-A2.5B-Thinking https://huggingface.co/JetBrains/Mellum2.1-12B-A2.5B-Thinking
- ByteDance-Seed/Seed-Coder-8B-Instruct , Hugging Face huggingface.co https://huggingface.co/ByteDance-Seed/Seed-Coder-8B-Instruct
- deepseek-ai/DeepSeek-V4-Pro , Hugging Face huggingface.co https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro
- LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Code arxiv.org https://arxiv.org/abs/2403.07974
- LiveCodeBench Leaderboard: Solar Pro 4 Leads at 87.8% benchlm.ai https://www.benchlm.ai/benchmarks/livecodebench
- Best LLM for Coding (October 2026): SWE-bench & LiveCodeBench Ranked benchlm.ai https://benchlm.ai/coding
Scores at a Glance
Score in %
Higher is better for this unit.
1
DeepSeek-V4-ProDeepSeekDeepSeek-V4-FlashDeepSeek
Claude Fable 5.1AnthropicVals.ai, max effort
Qwen3.8-27BAlibaba (Qwen)LiveCodeBench v6
Mellum2.1-12B-A2.5B-ThinkingJetBrainsthinking mode, LiveCodeBench v6QwQ-32B-PreviewAlibaba (Qwen)
Seed-Coder-8B-InstructByteDanceLiveCodeBench 2410-2502
DeepSeek-V4-ProDeepSeek93.5 %
vendor-reported2DeepSeek-V4-FlashDeepSeek91.6 %
vendor-reported3Claude Fable 5.1AnthropicVals.ai, max effort90.52 %
independent4Qwen3.8-27BAlibaba (Qwen)LiveCodeBench v690.3 %
vendor-reported5Mellum2.1-12B-A2.5B-ThinkingJetBrainsthinking mode, LiveCodeBench v682 %
vendor-reported6QwQ-32B-PreviewAlibaba (Qwen)50 %
vendor-reported7Seed-Coder-8B-InstructByteDanceLiveCodeBench 2410-250224.7 %
vendor-reported0255075100
Where this comes from
Each bar is one model, with its most independent recorded score. Hatched bars are vendor-reported; the word beside each bar says who ran it. Settings and dates differ between rows; the Scores table below has every result.
Scores
| Model | Score | Who ran it | Setting | View or population | Date | Source |
|---|---|---|---|---|---|---|
| DeepSeek-V4-Pro | 93.5 % | vendor-reported | – | – | – | blog.buildfastwithai.com |
| DeepSeek-V4-Flash | 91.6 % | vendor-reported | – | – | – | blog.buildfastwithai.com |
| Claude Fable 5.1 | 90.52 % ±0.86 | independent | Vals.ai, max effort | rank 1 of 143, – | – | vals.ai |
| Claude Fable 5.1 | 90.5 % | independent | Vals-run variant | rank 1, Vals.ai run shown on BenchLM | 10 Oct 2026 | benchlm.ai |
| Qwen3.8-27B | 90.3 % | vendor-reported | LiveCodeBench v6 | – | – | huggingface.co |
 Solar Pro 4 | 87.8 % | vendor-reported | – | rank 1, BenchLM main table | 10 Oct 2026 | benchlm.ai |
| Mellum2.1-12B-A2.5B-Thinking | 82 % | vendor-reported | thinking mode, LiveCodeBench v6 | – | – | huggingface.co |
| DeepSeek-V3.1-Terminus | 74.9 % | vendor-reported | Model card comparison table; mode not stated | DeepSeek-V3.1-Terminus model card | 22 Sep 2025 | huggingface.co |
| DeepSeek-V3.1 | 74.8 % | vendor-reported | Thinking mode, 2408 to 2505, Pass@1 | DeepSeek-V3.1 model card | 21 Aug 2025 | huggingface.co |
 Llama 3.1 Nemotron Ultra 253B v1 | 66.31 % | vendor-reported | Reasoning on, 2024-08-01 to 2025-02-01 window | Nemotron Ultra model card | Apr 2025 | huggingface.co |
| QwQ-32B-Preview | 50 % | vendor-reported | – | – | 28 Nov 2024 | qwenlm.github.io |
| DeepSeek-V2.5 (Dec 2024, DeepSeek-V2.5-1210) | 34.38 % | vendor-reported | 2024-08-01 to 2024-12-01 window; up from 29.2 | DeepSeek-V2.5-1210 model card | Dec 2024 | huggingface.co |
| Seed-Coder-8B-Instruct | 24.7 % | vendor-reported | LiveCodeBench 2410-2502 | – | – | huggingface.co |
Where this comes from
Scores from different settings, harnesses or dates are not directly comparable; the row says which. Sorted by score, highest first, as a reading aid, not a ranking.
To cite this page: Project AURION, AI Use-Cases Research & Innovation, "LiveCodeBench", https://aurion.semai.id/benchmarks/livecodebench/, records of October 2026.