// BENCHMARK

LiveCodeBench v6 benchmark

AI model leaderboard for the LiveCodeBench v6 benchmark. Compare how large language models score on LiveCodeBench v6, see the full ranking, and understand what this AI benchmark measures. Kimi K2 Thinking currently leads with 83.1%. LiveCodeBench v6 — newer cutoff continuing the LCB approach.

Leaderboard

# Model Organization Score Variant Source
#1 Kimi K2 Thinking Moonshot 83.1% thinking-no-tools official ↗
#2 GPT-5 OpenAI 80.6% thinking-high-cited-ring-1t official ↗
#3 Ring-1T Ant Group 78.3% thinking-avg@4 official ↗
#4 Qwen3-235B-A22B-Thinking (Jul 2025) Qwen 75.72% thinking-cited-ring-1t official ↗
#5 DeepSeek-V3.1-Terminus DeepSeek 75.33% thinking-cited-ring-1t official ↗
#6 gemini-2.5-pro Google DeepMind 70.65% cited-ring-1t official ↗
#7 DeepSeek-R1-0528 DeepSeek 70.3% reasoning official ↗
#8 EXAONE 4.0 (32B) LG AI Research 66.7% reasoning official ↗
#9 Qwen3-32B Qwen 60.1% reasoning official ↗
#10 Qwen3-235B-A22B-Thinking (Jul 2025) Qwen 58.9% reasoning official ↗
#11 Kimi K2 Moonshot 56.1% k2-0905-no-tools-cited-k2-thinking official ↗
#12 Qwen3-Next-80B-A3B Qwen 51.8% from-hf-readme official ↗
#13 Magistral Medium 1.1 Mistral AI 50.3% official ↗
#14 Magistral Small 1.1 Mistral AI 47.4% official ↗
#15 Phi-4-Reasoning-plus Microsoft 47.1% reasoning official ↗
#16 EXAONE 4.0 (1.2B) LG AI Research 45.3% reasoning official ↗
#17 DeepSeek-V3-0324 DeepSeek 44% official ↗
#18 Claude 3.5 Sonnet Anthropic 37.2% 1022-cited-mimo official ↗
#19 Llama 4 Maverick Meta AI 32.7% official ↗
#20 DeepSeek-R1-Distill-Qwen-14B Qwen 31.9% cited-mimo official ↗
#21 DeepSeek-R1-Distill-Qwen-14B DeepSeek 31.9% cited-mimo official ↗
#22 Mistral Medium 3 Mistral AI 30% cited-magistral official ↗
#23 Gemma 3 27B Google DeepMind 29.7% official ↗
#24 MiMo-7B-Base Xiaomi Corp 29.1% base official ↗
#25 Kimi Linear Moonshot 26% sft-pass@1 official ↗
#26 DeepSeek-R1-Distill-Qwen-7B DeepSeek 23.9% cited-mimo official ↗
#27 DeepSeek-R1-Distill-Qwen-7B Qwen 23.9% cited-mimo official ↗
#28 Qwen3-1.7B Qwen 16.6% non-reasoning official ↗
#29 Qwen3-0.6B Qwen 6.9% non-reasoning official ↗
#30 Gemma 3 1B Google DeepMind 2.3% non-reasoning official ↗

Frequently asked questions about LiveCodeBench v6

What is the LiveCodeBench v6 benchmark?

LiveCodeBench v6 — newer cutoff continuing the LCB approach.

How is the LiveCodeBench v6 benchmark scored?

LiveCodeBench v6 is scored using the pass@1 (%) metric, where a higher score is better. The maximum achievable score is 100.000. GenAIList aggregates reported scores from model providers and papers into a single ranked leaderboard.

Which AI model scores highest on LiveCodeBench v6?

As of the latest reported scores on GenAIList, Kimi K2 Thinking achieves the highest result on LiveCodeBench v6 with a score of 83.1%.

Is a higher LiveCodeBench v6 score better?

Yes. On LiveCodeBench v6 a higher score indicates better performance, so models near the top of the leaderboard are the strongest.