💻

LiveCodeBench

Live Code Bench

Continuously updated competitive programming problems

💻 Coding

Best: 91.7%

20 Models Evaluated

Champion

DeepSeek V4 Pro (Max)

DeepSeek

93.5%

Best Score

All models ranked by score on LiveCodeBench. Scores are relative to the best score of 91.7%.

#	Model	Provider	Score	vs Best
1	DeepSeek V4 Pro (Max)	DeepSeek	93.5%	Best
2	Step-3.5-Flash Elo 1389	Stepfun	86.4%	-5.3%
3	Kimi K2.5 Elo 1438	Moonshot	85%	-6.7%
4	GPT-5.5 Elo 1478	OpenAI	84%	-7.7%
5	Qwen 3.5 Elo 1450	Alibaba	83.6%	-8.1%
6	Kimi K2 Thinking	Moonshot	83.1%	-8.6%
7	Mistral Large Elo 1416	Mistral	82.8%	-8.9%
8	Gemini 3.1 Pro Elo 1492	Google	81.3%	-10.4%
9	MiMo-V2-Flash Elo 1393	Xiaomi	80.6%	-11.1%
10	Claude Opus 4.7 Elo 1500	Anthropic	80%	-11.7%
11	Grok 3 Elo 1412	xAI	79.4%	-12.3%
12	Claude Opus 4.6 Elo 1503	Anthropic	76%	-15.7%
13	DeepSeek V3.2 Elo 1423	DeepSeek	74.1%	-17.6%
14	Claude Sonnet 4.6 Elo 1460	Anthropic	72.4%	-19.3%
15	Gemini 3 Pro Elo 1486	Google	71%	-20.7%
16	Nemotron Ultra 253B Elo 1348	NVIDIA	66.3%	-25.4%
17	DeepSeek R1 Elo 1398	DeepSeek	65.9%	-25.8%
18	MiniMax M2.5 Elo 1404	MiniMax	65%	-26.7%
19	GPT-oss 120B Elo 1355	OpenAI	60%	-31.7%
20	GLM-5 Elo 1454	Zhipu AI	52%	-39.7%

DeepSeek V4 Pro (Max)

93.5%

Step-3.5-Flash

86.4%

Kimi K2.5

85%

GPT-5.5

84%

Qwen 3.5

83.6%

Kimi K2 Thinking

83.1%

Mistral Large

82.8%

Gemini 3.1 Pro

81.3%

MiMo-V2-Flash

80.6%

Claude Opus 4.7

80%

Learn how to evaluate your own models on LiveCodeBench.

LiveCodeBench is a standardized benchmark for evaluating AI model capabilities.

To run this benchmark on your own models, visit the official benchmark page or leaderboard for detailed instructions, code, and evaluation scripts.

Data compiled from official model cards, public leaderboards, and independent evaluations.