CL-bench leaderboard
- GPT-5.4 leads
- 27.9%
- models measured
- 18
- lead over #2
- +4.2 pts
- median
- 19.0%
Leader on CL-bench · read October 22
On October 22, 2026, 18 of our catalog's models had a CL-bench result, and OpenAI GPT-5.4 led with 27.9%, ahead of OpenAI GPT-5.1 at 23.7%.
Epoch AI collects its results; its benchmarking hub describes the tasks and links the authors.
Role in the index
CL-bench is shown as published and does not feed any of our indexes. Read the methodology
| # | Model | Result | Evidence | Coding index | Output $/Mtok |
|---|---|---|---|---|---|
| 1 | GPT-5.4OpenAI | 27.9% | 64.4 | $15.00 | |
| 2 | GPT-5.1OpenAI | 23.7% | 40.4 | $10.00 | |
| 3 | Grok 4.20xAI | 22.2% | 33.3 | $2.50 | |
| 4 | Claude Opus 4.5Anthropic | 21.1% | 44.0 | $25.00 | |
| 5 | Gemini 3.1 Pro PreviewGoogle | 20.8% | 32.3 | $12.00 | |
| 6 | Claude Opus 4.6Anthropic | 20.7% | 57.2 | $25.00 | |
| 7 | Qwen3.6 PlusQwen | 20.3% | 44.1 | $1.95 | |
| 8 | Qwen3.5 Plus 2026-04-20Qwen | 19.8% | 39.5 | $1.80 | |
| 9 | Kimi K2.5Moonshot AI | 19.3% | 38.6 | $2.85 | |
| 10 | GLM 5Z.ai | 18.7% | 41.7 | $3.20 | |
| 11 | GPT-5.2OpenAI | 18.2% | 55.9 | $14.00 | |
| 12 | o3OpenAI | 17.8% | Not benchmarked | $8.00 | |
| 13 | Kimi K2 ThinkingMoonshot AI | 17.6% | Not benchmarked | $2.50 | |
| 14 | GLM 4.7Z.ai | 15.9% | 55.1 | $2.20 | |
| 15 | DeepSeek V3.2 ExpDeepSeek | 13.2% | Not benchmarked | $0.41 | |
| 16 | DeepSeek V3.2DeepSeek | 12.4% | 30.0 | $0.88 | |
| 17 | Kimi K2 0711Moonshot AI | 11.9% | Not benchmarked | $2.30 | |
| 18 | MiniMax M2.5MiniMax | 11.4% | 37.2 | $1.20 |
Only models in our catalog are listed, and each appears once with its best published run. 215 rows were read from the source on October 22, 2026, and rows that match no catalog model are left out.