CLBench - Procedural Task Execution: leaderboard
CL-bench (context learning) category 'Procedural Task Execution': tasks that require executing a procedure supplied in a professional context the model has to learn from. Solving rate in percent (published as a 0-1 fraction), higher is better; labels keep the reasoning effort the source records.
Metric: Solving Rate (%). Source: www.clbench.com. Saturation forecast: Around 2030. 36 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.4 (xHigh) | 30.1 | #83 (GPT-5.4) |
| 2 | Claude Opus 4.6 | 24.4 | #67 |
| 3 | Gemini 3.1 Pro (Preview) (High) | 23.8 | #64 (Gemini 3.1 Pro (Preview)) |
| 4 | GPT-5.1 (High) | 23.8 | #145 (GPT-5.1) |
| 5 | Grok 4.20 (Reasoning) | 23.4 | #179 (Grok 4.20) |
| 6 | GPT-5.1 | 22.8 | #145 |
| 7 | Claude Opus 4.5 (Thinking) | 22.6 | #87 (Claude Opus 4.5) |
| 8 | Hy3-preview | 22.3 | #166 |
| 9 | GLM-5 | 21.6 | #149 |
| 10 | GPT-5.2 (High) | 21.4 | #115 (GPT-5.2) |
| 11 | Seed 2.0 Pro (High) | 21.4 | #122 (Seed 2.0 Pro) |
| 12 | Qwen 3.6 Plus | 21.3 | #124 |
| 13 | Kimi K2.5 | 19.5 | #152 |
| 14 | Claude Opus 4.5 | 19.5 | #87 |
| 15 | O3 (High) | 19.5 | #132 (O3) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=clbench-procedural-task-execution · How It Works · Data refreshed daily, snapshot 2026-11-05.