CL-bench Life
Rank #30 of 51 · 13 models scored · top score 22.2
unknown
Strongest on contamination resistance, weakest on discrimination. Caveat: the public test set leaves contamination on the table.
Weighted composite. Each signal is computed from the source-backed leaderboard below.
Does it still separate models?
How far from ceiling / clustered at the top?
Public vs held-out; training-leak risk.
Apples-to-apples, or mixed / vendor-optimized?
How old is the benchmark?
| # | Model | Score | Evidence | Measured |
|---|---|---|---|---|
| 1 | GPT-5.5 | 22.2 | unverified | 2026-04-23 |
| 2 | GPT-5.4 | 21.7 | unverified | 2026-03-05 |
| 3 | GPT-5.1 | 17.3 | unverified | 2025-11-13 |
| 4 | Claude Opus 4.6 | 17 | unverified | 2026-02-05 |
| 5 | Gemini 3.1 Pro | 16.9 | unverified | 2026-02-19 |
| 6 | DeepSeek-V4-Pro | 13.5 | unverified | 2026-04-24 |
| 7 | Kimi K2.5 | 13.2 | unverified | 2026-02-02 |
| 8 | Qwen 3.5 Plus (hosted 397B-A17B) | 12.4 | unverified | 2026-02-16 |
| 9 | Grok 4.20 | 11.9 | unverified | 2026-02-17 |
| 10 | GLM-4.7 | 10.9 | unverified | 2025-12-22 |
| 11 | DeepSeek-V3.2-Exp | 9.5 | unverified | 2025-09-29 |
| 12 | DeepSeek-V3.2 | 7.4 | unverified | 2025-12-01 |
| 13 | MiniMax-M2.5 | 6.3 | unverified | 2026-02-12 |