MTAC-IFBench - CSR (OpenCode): leaderboard
Multi-turn agentic coding instruction following under the OpenCode framework: Constraint Success Rate, the share of user constraints each turn's response satisfies, averaged over turn buckets 1-10. Higher is better. Source: arxiv.org/abs/2609.14992.
Metric: Constraint Success Rate (%, average over turn buckets). Source: arxiv.org. Saturation forecast: Around February 2027. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GLM-5.2 | 76.8 | #80 |
| 2 | GLM-5.1 | 74.3 | #109 |
| 3 | DeepSeek V4 Pro | 72.2 | #112 |
| 4 | Qwen 3.6 Plus | 68 | #124 |
| 5 | Seed 2.0 Pro | 67.2 | #122 |
| 6 | DeepSeek V4 Flash | 63.7 | #155 |
| 7 | Claude Haiku 4.5 | 61.8 | #287 |
| 8 | Qwen 3.5 27B | 61.8 | #192 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=mtac-ifbench-csr-opencode · How It Works · Data refreshed daily, snapshot 2026-11-05.