LiveCodeBench — Specific CoT-prompted Pass@1 metric variant, distinct from plain pass@1.
| # | Model | Vendor | Best score | Runs | Last seen |
|---|---|---|---|---|---|
| 1 | DeepSeek-R1 | DeepSeek | 65.9 | 6 | 2026-06-04 |
| 2 | o1 | OpenAI | 63.4 | 5 | 2026-06-04 |
| 3 | OpenAI o1-mini | OpenAI | 53.8 | 5 | 2026-06-04 |
| 4 | R1-Zero | DeepSeek | 50 | 1 | 2026-05-03 |
| 5 | DeepSeek-V3 | DeepSeek | 40.5 | 1 | 2026-05-03 |
| 6 | Claude 3.5 Sonnet | Anthropic | 36.3 | 6 | 2026-06-04 |
| 7 | GPT-4o | OpenAI | 34.2 | 6 | 2026-06-04 |
| 8 | Qwen2.5 Instruct 72B | Alibaba | 31.1 | 1 | 2026-05-03 |
| 9 | DeepSeek-V2.5 | DeepSeek | 29.2 | 1 | 2026-05-03 |
| 10 | Llama 3.1 Instruct 405B | Meta | 28.4 | 1 | 2026-05-03 |
| 11 | DeepSeek-V2 | DeepSeek | 18.8 | 1 | 2026-05-03 |
Best tracked score per model (default configuration; source-attributed and verification-tiered). Open a model for its full benchmark surface, provenance and pricing.