CL-bench
Each model’s result as Epoch AI’s Benchmarking Hub publishes it, run by Epoch or collected from the source it cites. Facthory has not run them.
Data: Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks, licensed CC BY 4.0. Read on 16 Nov 2026.
| # | Model | Organization | Result | Evidence |
|---|---|---|---|---|
| 1 | GPT-5.4 (xhigh) API access | OpenAI | 27.9% | – |
| 2 | GPT-5.1 (high) API access | OpenAI | 23.7% | – |
| 3 | grok-4-20 API access | xAI | 22.2% | – |
| 4 | Claude Opus 4.5 (unknown thinking) API access | Anthropic | 21.1% | – |
| 5 | GPT-5.1 (unknown thinking) API access | OpenAI | 21.1% | – |
| 6 | Gemini 3.1 Pro Preview API access | Google DeepMind | 20.8% | – |
| 7 | Claude Opus 4.6 (unknown thinking) API access | Anthropic | 20.7% | – |
| 8 | Qwen 3.6 Plus (2026-04-02) API access | Alibaba | 20.3% | – |
| 9 | Qwen3.5 Plus | Alibaba | 19.8% | – |
| 10 | Kimi K2.5 Open weights (unrestricted) | Moonshot | 19.3% | – |
| 11 | GLM-5 Open weights (unrestricted) | Z.ai (Zhipu AI) | 18.7% | – |
| 12 | GPT-5.2 (unknown thinking) API access | OpenAI | 18.2% | – |
| 13 | GPT-5.2 (high) API access | OpenAI | 18.1% | – |
| 14 | o3 (high) API access | OpenAI | 17.8% | – |
| 15 | Kimi K2 Thinking (Together) Open weights (restricted use) | Moonshot | 17.6% | – |
| 16 | GLM-4.7 (Novita) Open weights (unrestricted) | Z.ai (Zhipu AI) | 15.9% | – |
| 17 | Gemini 3 Pro Preview API access | Google DeepMind | 15.8% | – |
| 18 | mimo-v2-pro API access | Xiaomi Corp | 15.7% | – |
| 19 | Qwen3-Max-Instruct API access | Alibaba | 14.5% | – |
| 20 | DeepSeek-V3.2-Exp_thinking Open weights (unrestricted) | DeepSeek | 13.2% | – |
| 21 | DeepSeek-V3.2 (Thinking; Novita) Open weights (unrestricted) | DeepSeek | 12.4% | – |
| 22 | Kimi K2 Thinking Open weights (restricted use) | Moonshot | 11.9% | – |
| 23 | MiniMax-M2.5 Open weights (unrestricted) | MiniMax | 11.4% | – |
