SQBench - L2 Composite Skills: leaderboard
Metric: Strict Pass@1 (%; the 60 L2 composite-skill tasks on SQBench v1.0, 220 production-oriented delivery tasks in which an agent processes input assets, uses tools and produces a specified deliverable; a task passes only with functional Completion 1 and no triggered 10-dimension risk penalty; one run per configuration and task; higher is better). Source: arxiv.org. Saturation forecast: Around December 2026. 27 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Kimi K3 | 71.7 |
| 2 | Claude Opus 4.8 (Max) | 66.7 |
| 3 | Kimi K2.7 Code | 63.3 |
| 4 | Qwen 3.7 Max | 61.7 |
| 5 | GLM-5.2 (Max) | 61.7 |
| 6 | Doubao-Seed-2.1-Pro (High) | 61.7 |
| 7 | MiniMax-M3 | 60 |
| 8 | MiMo-V2.5 | 60 |
| 9 | DeepSeek V4 Pro (Max) | 60 |
| 10 | DeepSeek V4 Pro (High) | 60 |
| 11 | DeepSeek V4 Flash (High) | 60 |
| 12 | Kimi K2.6 | 58.3 |
| 13 | Qwen 3.6 Plus | 56.7 |
| 14 | GLM-5.1 | 56.7 |
| 15 | MiMo-V2.5-Pro | 56.7 |
Interactive version: theaggregate.ai/benchmark?slug=sqbench-l2-composite-skills · How It Works · Data refreshed daily, snapshot 2026-09-29.