C3-Bench (Controllable Code Completion) - Implementation Control: leaderboard
Metric: Instruction-Following Rate (%; 1,286 ICC tasks: passes the unit tests and a Claude3.5-Sonnet judge finds the instructed implementation). Source: arxiv.org. Saturation forecast: Estimated already saturated. 42 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | O1 (2024-12-17) | 62.9 |
| 2 | Claude 3.5 Sonnet (20241022) | 60.9 |
| 3 | GPT-4o (2024-11-20) | 59.6 |
| 4 | Gemini 2.0 Flash | 59.5 |
| 5 | GPT-4o (2024-08-06) | 58.6 |
| 6 | O1 Preview | 57.7 |
| 7 | O1 Mini | 55 |
| 8 | DeepSeek V3 (0324) | 53 |
| 9 | DeepSeek V3 | 47.3 |
| 10 | GPT-4o Mini (2024-07-18) | 42.2 |
| 11 | GPT-4o (2024-05-13) | 41.3 |
| 12 | Claude 3.5 Haiku | 40.7 |
| 13 | Qwen 2.5 Coder 32B | 38.7 |
| 14 | Qwen 2.5 Coder 14B | 36.2 |
| 15 | Codestral-22B-v0.1 | 34.1 |
Interactive version: theaggregate.ai/benchmark?slug=c3-bench-controllable-code-completion-implementation-control · How It Works · Data refreshed daily, snapshot 2026-09-25.