C3-Bench (Controllable Code Completion) - Scale Control: leaderboard
Metric: Instruction-Following Rate (%; 909 SCC tasks: AST node-type and line-count check of the requested completion scope). Source: arxiv.org. Saturation forecast: Estimated already saturated. 42 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | O1 (2024-12-17) | 59.6 |
| 2 | Claude 3.5 Sonnet (20241022) | 50.8 |
| 3 | O1 Preview | 48.9 |
| 4 | O1 Mini | 44.7 |
| 5 | Claude 3.5 Haiku | 26.4 |
| 6 | GPT-4o (2024-08-06) | 24.1 |
| 7 | GPT-4o (2024-11-20) | 24.1 |
| 8 | GPT-4o Mini (2024-07-18) | 22.5 |
| 9 | GPT-4o (2024-05-13) | 20.8 |
| 10 | DeepSeek V3 | 20.2 |
| 11 | DeepSeek V3 (0324) | 20.2 |
| 12 | Qwen 2.5 Coder 32B Instruct | 16.9 |
| 13 | Qwen 2.5 Coder 14B Instruct | 13.5 |
| 14 | Qwen 2.5 72B Instruct | 9.4 |
| 15 | Qwen 2.5 Coder 14B | 8.7 |
Interactive version: theaggregate.ai/benchmark?slug=c3-bench-controllable-code-completion-scale-control · How It Works · Data refreshed daily, snapshot 2026-09-25.