CL-bench
ReportedNo description yet: we only describe an eval set when we can do it correctly.
- Kind
- Published by a third party's leaderboard, collected by Epoch AI. The leaderboard is named and linked.
- Score shown
- Epoch's column "Overall", as a percentage, higher is better; random guessing scores about 0.0%. The source gives no error bars.
- Results
- 23 in all: 8 for open-weight models, 14 for closed ones, 1 where Epoch does not say. 8 linked to a model page here.
Best open-weight result: moonshotai/Kimi-K2.5, 19.3%, number 10 of 23 results.
| # | Model | Weights | Setting | Score | Date | Source |
|---|---|---|---|---|---|---|
| 1 | gpt-5.4-2026-03-05_xhighOpenAI | closed | effort xhigh | 27.9% | not dated | via Epoch AI |
| 2 | gpt-5.1-2025-11-13_highOpenAI | closed | effort high | 23.7% | not dated | via Epoch AI |
| 3 | grok-4-20xAI | closed | default | 22.2% | not dated | via Epoch AI |
| 4 | claude-opus-4-5-20251101_unknownAnthropic | closed | effort unknown | 21.1% | not dated | via Epoch AI |
| 5 | gpt-5.1-2025-11-13_unknownOpenAI | closed | effort unknown | 21.1% | not dated | via Epoch AI |
| 6 | gemini-3.1-pro-previewGoogle DeepMind | closed | default | 20.8% | not dated | via Epoch AI |
| 7 | claude-opus-4-6_unknownAnthropic | closed | effort unknown | 20.7% | not dated | via Epoch AI |
| 8 | qwen3.6-plusAlibaba | closed | default | 20.3% | not dated | via Epoch AI |
| 9 | qwen3.5-plusAlibaba | not stated | default | 19.8% | not dated | via Epoch AI |
| 10 | moonshotai/Kimi-K2.5Epoch: kimi-k2.5, Moonshot | open | default | 19.3% | not dated | via Epoch AI |
| 11 | zai-org/GLM-5Epoch: glm-5, Z.ai (Zhipu AI) | open | default | 18.7% | not dated | via Epoch AI |
| 12 | gpt-5.2-2025-12-11_unknownOpenAI | closed | effort unknown | 18.2% | not dated | via Epoch AI |
| 13 | gpt-5.2-2025-12-11_highOpenAI | closed | effort high | 18.1% | not dated | via Epoch AI |
| 14 | o3-2025-04-16_highOpenAI | closed | effort high | 17.8% | not dated | via Epoch AI |
| 15 | moonshotai/Kimi-K2-ThinkingEpoch: moonshotai/Kimi-K2-Thinking, Moonshot | open | default | 17.6% | not dated | via Epoch AI |
| 16 | zai-org/GLM-4.7Epoch: zai-org/glm-4.7, Z.ai (Zhipu AI) | open | default | 15.9% | not dated | via Epoch AI |
| 17 | gemini-3-pro-previewGoogle DeepMind | closed | default | 15.8% | not dated | via Epoch AI |
| 18 | mimo-v2-proXiaomi Corp | closed | default | 15.7% | not dated | via Epoch AI |
| 19 | qwen3-max-2025-09-23Alibaba | closed | default | 14.5% | not dated | via Epoch AI |
| 20 | deepseek-ai/DeepSeek-V3.2-ExpEpoch: DeepSeek-V3.2-Exp_thinking, DeepSeek | open | effort thinking | 13.2% | not dated | via Epoch AI |
| 21 | deepseek-ai/DeepSeek-V3.2Epoch: deepseek/deepseek-v3.2, DeepSeek | open | default | 12.4% | not dated | via Epoch AI |
| 22 | moonshotai/Kimi-K2-ThinkingEpoch: kimi-k2-thinking, Moonshot | open | default | 11.9% | not dated | via Epoch AI |
| 23 | MiniMaxAI/MiniMax-M2.5Epoch: MiniMax-M2.5, MiniMax | open | default | 11.4% | not dated | via Epoch AI |
Open-weight rows are shaded. Dates are when the result was run or evaluated, where the data says. Data from Epoch AI (CC-BY 4.0), fetched 19 Nov 2026: Epoch AI, ‘Capabilities & benchmarking’. Published online at epoch.ai. Retrieved from ‘https://epoch.ai/benchmarks’ [online resource]. Epoch AI. Method