CodeElo: leaderboard
CodeElo — competitive-programming benchmark that rates LLMs with human-comparable Elo using CodeForces contest problems and official judging.
Metric: Elo Rating. Source: codeelo-bench.github.io. Saturation forecast: Estimated already saturated. 33 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Best Expert | 2157 | |
| 2 | O1 Mini | 1578 | #323 |
| 3 | QwQ 32B-Preview | 1261 | #651 |
| 4 | Median Expert | 1073 | |
| 5 | Qwen 2.5 72B Instruct | 634 | #454 |
| 6 | Mistral Large 2 (Nov) Instruct (2411) | 631 | #465 |
| 7 | Qwen 2.5 Coder 32B Instruct | 575 | #528 |
| 8 | Qwen 2.5 32B Instruct | 513 | #498 |
| 9 | Llama 3.1 70B Instruct | 478 | #561 |
| 10 | Qwen 2.5 Coder 14B Instruct | 424 | #675 |
| 11 | Qwen 2.5 14B Instruct | 414 | #647 |
| 12 | Qwen 2.5 Coder 7B Instruct | 397 | #951 |
| 13 | Codestral-22B-v0.1 | 385 | #911 |
| 14 | Qwen 2.5 7B Instruct | 315 | #862 |
| 15 | Yi-Coder-9B-Chat | 296 | #1262 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=codeelo · How It Works · Data refreshed daily, snapshot 2026-10-20.