Atlas

Benchmarks

← All benchmarks

LiveCodeBench - August 2024 to May 2025

Code

LiveCodeBench code-generation problems in the August 1, 2024 to May 1, 2025 date window (the v6 cohort), labelled 2408-2505 in DeepSeek's R1-0528, V3.1 and V3.2-Exp cards, 2024.08-2025.04 in its V3.2 report, and "Aug 24 - May 25" in Moonshot's Kimi K2 report and NVIDIA's Nemotron cards. This exact historical problem cohort is retained as a diagnostic rather than pooled with other LiveCodeBench windows.

Top models (higher is better)

ModelScore
DeepSeek-V4-Flash-073194.6
Qwen3.8-Flash-Next94.1
Gemini 3 Pro Preview90.7
Nemotron 3 Ultra 550B A55B89.2
DeepSeek-V3.2-Speciale88.7
Nemotron Cascade 2 30B A3B88.4
Inkling-Small88.3
Nemotron 3.5 Super VL 120B A12B BF1688.2
Nemotron 3.5 Super VL 120B A12B NVFP487.4
GPT-584.5
DeepSeek-V3.283.3
MiniMax M283.0
Kimi K2 Thinking82.6
GLM-5.3 Flash80.6
o377.3
Qwen3.5 122B A10B76.7
DeepSeek-V3.1-Terminus74.9
DeepSeek-V3.174.8
Nemotron Cascade 14B Thinking74.6
DeepSeek-V3.2-Exp74.1
DeepSeek-R1-052873.3
Gemini 2.5 Pro Preview 05-0671.8
Nemotron Cascade 8B Thinking71.4
Nemotron Cascade 8B71.1
Qwen3-235B-A22B66.5
o3-mini65.9
Nemotron Nano 9B V265.3
Sonnet 4.564.0
DeepSeek-R163.5
Qwen3 14B63.5
Gemini 2.5 Flash62.3
DeepSeek R1 0528 Qwen3 8B60.5
Qwen3 8B58.3
Kimi K2 Instruct53.7
Sonnet 448.5
Opus 447.4
GPT-4.144.7
Loading Atlas data…