Lech Mazur Writing
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.
- Link
- https://github.com/lechmazur/Writing
- Score shown
- Epoch's column "Mean score", as a percentage, higher is better. The source gives no error bars.
- Results
- 49 in all: 14 for open-weight models, 35 for closed ones. 13 linked to a model page here.
Best open-weight result: moonshotai/Kimi-K2-Instruct, 85.6%, number 2 of 49 results.
| # | Model | Weights | Setting | Score | Date | Source |
|---|---|---|---|---|---|---|
| 1 | gpt-5-2025-08-07_mediumOpenAI | closed | effort medium | 86.0% | not dated | lechmazur/writing Github repository |
| 2 | moonshotai/Kimi-K2-InstructEpoch: Kimi-K2-Instruct, Moonshot | open | default | 85.6% | not dated | lechmazur/writing Github repository |
| 3 | claude-opus-4-1-20250805Anthropic | closed | default | 84.7% | not dated | lechmazur/writing Github repository |
| 4 | claude-opus-4-1-20250805_16KAnthropic | closed | default | 84.5% | not dated | lechmazur/writing Github repository |
| 5 | o3-pro-2025-06-10_mediumOpenAI | closed | effort medium | 84.4% | not dated | lechmazur/writing Github repository |
| 6 | o3-2025-04-16_mediumOpenAI | closed | effort medium | 83.9% | not dated | lechmazur/writing Github repository |
| 7 | gemini-2.5-proGoogle DeepMind | closed | default | 83.8% | not dated | lechmazur/writing Github repository |
| 8 | claude-opus-4-20250514_16KAnthropic | closed | default | 83.6% | not dated | lechmazur/writing Github repository |
| 9 | claude-opus-4-20250514Anthropic | closed | default | 83.1% | not dated | lechmazur/writing Github repository |
| 10 | gpt-5-mini-2025-08-07_mediumOpenAI | closed | effort medium | 83.1% | not dated | lechmazur/writing Github repository |
| 11 | deepseek-ai/DeepSeek-R1Epoch: DeepSeek-R1, DeepSeek | open | default | 83.0% | not dated | lechmazur/writing Github repository |
| 12 | Qwen/Qwen3-235B-A22BEpoch: qwen3-235b-a22b, Alibaba | open | default | 83.0% | not dated | lechmazur/writing Github repository |
| 13 | Qwen/Qwen3-235B-A22B-Thinking-2507Epoch: Qwen3-235B-A22B-Thinking-2507, Alibaba | open | default | 82.4% | not dated | lechmazur/writing Github repository |
| 14 | deepseek-ai/DeepSeek-R1-0528Epoch: DeepSeek-R1-0528, DeepSeek | open | default | 81.9% | not dated | lechmazur/writing Github repository |
| 15 | gpt-4o-2024-11-20OpenAI | closed | default | 81.8% | not dated | lechmazur/writing Github repository |
| 16 | claude-sonnet-4-20250514_16KAnthropic | closed | default | 81.4% | not dated | lechmazur/writing Github repository |
| 17 | claude-3-7-sonnet-20250219_16KAnthropic | closed | default | 81.1% | not dated | lechmazur/writing Github repository |
| 18 | claude-sonnet-4-20250514Anthropic | closed | default | 80.9% | not dated | lechmazur/writing Github repository |
| 19 | gemini-2.5-pro-preview-05-06Google DeepMind | closed | default | 80.9% | not dated | lechmazur/writing Github repository |
| 20 | gemini-2.5-pro-exp-03-25Google DeepMind | closed | default | 80.5% | not dated | lechmazur/writing Github repository |
| 21 | claude-3-5-sonnet-20241022Anthropic | closed | default | 80.3% | not dated | lechmazur/writing Github repository |
| 22 | Qwen/QwQ-32BEpoch: QwQ-32B (16K thinking), Alibaba | open | default | 80.2% | not dated | lechmazur/writing Github repository |
| 23 | google/gemma-3-27b-itEpoch: gemma-3-27b-it, Google DeepMind | open | default | 79.9% | not dated | lechmazur/writing Github repository |
| 24 | claude-3-7-sonnet-20250219Anthropic | closed | default | 79.4% | not dated | lechmazur/writing Github repository |
| 25 | mistral-medium-2505Mistral AI | closed | default | 77.3% | not dated | lechmazur/writing Github repository |
| 26 | openai/gpt-oss-120bEpoch: gpt-oss-120b, OpenAI | open | default | 77.1% | not dated | lechmazur/writing Github repository |
| 27 | deepseek-ai/DeepSeek-V3-0324Epoch: DeepSeek-V3-0324, DeepSeek | open | default | 77.0% | not dated | lechmazur/writing Github repository |
| 28 | grok-4-0709xAI | closed | default | 76.9% | not dated | lechmazur/writing Github repository |
| 29 | gemini-2.5-flash-preview-04-17 (24K thinking)Google DeepMind | closed | default | 76.5% | not dated | lechmazur/writing Github repository |
| 30 | grok-3-betaxAI | closed | default | 76.4% | not dated | lechmazur/writing Github repository |
| 31 | gpt-4.5-preview-2025-02-27OpenAI | closed | default | 75.6% | not dated | lechmazur/writing Github repository |
| 32 | Qwen/Qwen3-30B-A3BEpoch: Qwen3-30B-A3B, Alibaba | open | default | 75.3% | not dated | lechmazur/writing Github repository |
| 33 | o4-mini-2025-04-16_mediumOpenAI | closed | effort medium | 75.0% | not dated | lechmazur/writing Github repository |
| 34 | gemini-2.0-flash-thinking-exp-01-21Google DeepMind,Google | closed | default | 73.8% | not dated | lechmazur/writing Github repository |
| 35 | claude-3-5-haiku-20241022Anthropic | closed | default | 73.5% | not dated | lechmazur/writing Github repository |
| 36 | grok-3-mini-beta_lowxAI | closed | effort low | 73.5% | not dated | lechmazur/writing Github repository |
| 37 | zai-org/GLM-4.5Epoch: glm-4.5, Z.ai (Zhipu AI),Tsinghua University | open | default | 73.4% | not dated | lechmazur/writing Github repository |
| 38 | qwen2.5-maxAlibaba | closed | default | 72.9% | not dated | lechmazur/writing Github repository |
| 39 | gemini-2.0-flash-expGoogle DeepMind,Google | closed | default | 71.5% | not dated | lechmazur/writing Github repository |
| 40 | o1-2024-12-17_mediumOpenAI | closed | effort medium | 70.2% | not dated | lechmazur/writing Github repository |
| 41 | mistral-large-2407Mistral AIno page match | open | default | 69.0% | not dated | lechmazur/writing Github repository |
| 42 | gpt-4o-mini-2024-07-18OpenAI | closed | default | 67.2% | not dated | lechmazur/writing Github repository |
| 43 | o1-mini-2024-09-12_mediumOpenAI | closed | effort medium | 64.9% | not dated | lechmazur/writing Github repository |
| 44 | grok-2-1212xAI | closed | default | 63.6% | not dated | lechmazur/writing Github repository |
| 45 | microsoft/phi-4Epoch: phi-4, Microsoft Research | open | default | 62.6% | not dated | lechmazur/writing Github repository |
| 46 | meta-llama/Llama-4-Maverick-17B-128E-InstructEpoch: Llama-4-Maverick-17B-128E-Instruct, Meta AI | open | default | 62.0% | not dated | lechmazur/writing Github repository |
| 47 | o3-mini-2025-01-31_highOpenAI | closed | effort high | 61.7% | not dated | lechmazur/writing Github repository |
| 48 | o3-mini-2025-01-31_mediumOpenAI | closed | effort medium | 61.5% | not dated | lechmazur/writing Github repository |
| 49 | amazon.nova-pro-v1:0Amazon | closed | default | 60.5% | not dated | lechmazur/writing Github repository |
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