CLBench Life: leaderboard
CL-bench Life overall solving rate: context-learning tasks built on everyday, real-life contexts (conversations, fragmented notes, activity records) that the model has to learn from. Percent (published as a 0-1 fraction), higher is better; labels keep reasoning effort such as High, Low or NoThink.
Metric: Solving Rate (%). Source: www.clbench.com. Saturation forecast: Around 2029. 29 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 (High) | 22.2 | #33 (GPT-5.5) |
| 2 | GPT-5.4 (xHigh) | 21.7 | #83 (GPT-5.4) |
| 3 | GPT-5.4 (High) | 19.3 | #83 (GPT-5.4) |
| 4 | GPT-5.1 (High) | 17.3 | #145 (GPT-5.1) |
| 5 | Claude Opus 4.6 (High) | 17 | #67 (Claude Opus 4.6) |
| 6 | Gemini 3.1 Pro (Preview) (High) | 16.9 | #64 (Gemini 3.1 Pro (Preview)) |
| 7 | Gemini 3.1 Pro (Preview) | 16 | #64 |
| 8 | Hy3-preview (High) | 15.7 | #166 (Hy3-preview) |
| 9 | Seed 2.0 Pro (High) | 15.5 | #122 (Seed 2.0 Pro) |
| 10 | GPT-5.4 | 13.8 | #83 |
| 11 | Claude Opus 4.6 | 13.6 | #67 |
| 12 | DeepSeek V4 Pro (High) | 13.5 | #112 (DeepSeek V4 Pro) |
| 13 | Kimi K2.5 (High) | 13.2 | #152 (Kimi K2.5) |
| 14 | Grok 4.20 (High) | 11.9 | #179 (Grok 4.20) |
| 15 | Seed 2.0 Pro (Medium) | 11.4 | #122 (Seed 2.0 Pro) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=clbench-life · How It Works · Data refreshed daily, snapshot 2026-11-05.