Leaderboards
Benchmarks
Every score carries its origin.
LiveCodeBench
% pass on contamination-free competitive programming problemsProvenance16 modelsscores Sep 21, 2026 – Sep 28, 2026
1o4 Mini HighOpenAIScore87.3Score / $20ScoredSep 28, 20262o3OpenAIScore84.7Score / $11ScoredSep 28, 20263o4 MiniOpenAIScore84.5Score / $19ScoredSep 28, 20264R1 0528DeepSeek685BScore84.4Score / $39Score / GB VRAM~0.2ScoredSep 28, 20265Gemini 2.5 ProGoogleScore84.3Score / $17ScoredSep 28, 20266Qwen3 235B A22B Instruct 2507Qwen235BScore80.4Score / $230Score / GB VRAM~0.54ScoredSep 28, 20267o3 MiniOpenAIScore77.7Score / $18ScoredSep 28, 20268Gemini 2.5 FlashGoogleScore76.2Score / $61ScoredSep 28, 20269Claude Opus 4AnthropicScore62.4Score / $1ScoredSep 28, 202610Claude Sonnet 4AnthropicScore59.4Score / $4ScoredSep 28, 202611DeepSeek V3DeepSeek685BScore49.6Score / $56Score / GB VRAM0ScoredSep 28, 202612Claude Sonnet 4.5AnthropicScore48.7Score / $3ScoredSep 21, 202613GPT-4o (2024-08-06)OpenAIScore38.3Score / $4ScoredSep 28, 202614GPT-4 TurboOpenAIScore37.3Score / $1ScoredSep 28, 202615GPT-4o-mini (2024-07-18)OpenAIScore35.5Score / $59ScoredSep 28, 202616Claude 3 HaikuAnthropicScore22.5Score / $18ScoredSep 28, 2026