Skip to content
Weights Not Walls
Menu

Lech Mazur Writing

Reported

No 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.

Lech Mazur Writing results
#ModelWeightsSettingScoreDateSource
1gpt-5-2025-08-07_mediumOpenAIclosedeffort medium86.0%not datedlechmazur/writing Github repository
2moonshotai/Kimi-K2-InstructEpoch: Kimi-K2-Instruct, Moonshotopendefault85.6%not datedlechmazur/writing Github repository
3claude-opus-4-1-20250805Anthropiccloseddefault84.7%not datedlechmazur/writing Github repository
4claude-opus-4-1-20250805_16KAnthropiccloseddefault84.5%not datedlechmazur/writing Github repository
5o3-pro-2025-06-10_mediumOpenAIclosedeffort medium84.4%not datedlechmazur/writing Github repository
6o3-2025-04-16_mediumOpenAIclosedeffort medium83.9%not datedlechmazur/writing Github repository
7gemini-2.5-proGoogle DeepMindcloseddefault83.8%not datedlechmazur/writing Github repository
8claude-opus-4-20250514_16KAnthropiccloseddefault83.6%not datedlechmazur/writing Github repository
9claude-opus-4-20250514Anthropiccloseddefault83.1%not datedlechmazur/writing Github repository
10gpt-5-mini-2025-08-07_mediumOpenAIclosedeffort medium83.1%not datedlechmazur/writing Github repository
11deepseek-ai/DeepSeek-R1Epoch: DeepSeek-R1, DeepSeekopendefault83.0%not datedlechmazur/writing Github repository
12Qwen/Qwen3-235B-A22BEpoch: qwen3-235b-a22b, Alibabaopendefault83.0%not datedlechmazur/writing Github repository
13Qwen/Qwen3-235B-A22B-Thinking-2507Epoch: Qwen3-235B-A22B-Thinking-2507, Alibabaopendefault82.4%not datedlechmazur/writing Github repository
14deepseek-ai/DeepSeek-R1-0528Epoch: DeepSeek-R1-0528, DeepSeekopendefault81.9%not datedlechmazur/writing Github repository
15gpt-4o-2024-11-20OpenAIcloseddefault81.8%not datedlechmazur/writing Github repository
16claude-sonnet-4-20250514_16KAnthropiccloseddefault81.4%not datedlechmazur/writing Github repository
17claude-3-7-sonnet-20250219_16KAnthropiccloseddefault81.1%not datedlechmazur/writing Github repository
18claude-sonnet-4-20250514Anthropiccloseddefault80.9%not datedlechmazur/writing Github repository
19gemini-2.5-pro-preview-05-06Google DeepMindcloseddefault80.9%not datedlechmazur/writing Github repository
20gemini-2.5-pro-exp-03-25Google DeepMindcloseddefault80.5%not datedlechmazur/writing Github repository
21claude-3-5-sonnet-20241022Anthropiccloseddefault80.3%not datedlechmazur/writing Github repository
22Qwen/QwQ-32BEpoch: QwQ-32B (16K thinking), Alibabaopendefault80.2%not datedlechmazur/writing Github repository
23google/gemma-3-27b-itEpoch: gemma-3-27b-it, Google DeepMindopendefault79.9%not datedlechmazur/writing Github repository
24claude-3-7-sonnet-20250219Anthropiccloseddefault79.4%not datedlechmazur/writing Github repository
25mistral-medium-2505Mistral AIcloseddefault77.3%not datedlechmazur/writing Github repository
26openai/gpt-oss-120bEpoch: gpt-oss-120b, OpenAIopendefault77.1%not datedlechmazur/writing Github repository
27deepseek-ai/DeepSeek-V3-0324Epoch: DeepSeek-V3-0324, DeepSeekopendefault77.0%not datedlechmazur/writing Github repository
28grok-4-0709xAIcloseddefault76.9%not datedlechmazur/writing Github repository
29gemini-2.5-flash-preview-04-17 (24K thinking)Google DeepMindcloseddefault76.5%not datedlechmazur/writing Github repository
30grok-3-betaxAIcloseddefault76.4%not datedlechmazur/writing Github repository
31gpt-4.5-preview-2025-02-27OpenAIcloseddefault75.6%not datedlechmazur/writing Github repository
32Qwen/Qwen3-30B-A3BEpoch: Qwen3-30B-A3B, Alibabaopendefault75.3%not datedlechmazur/writing Github repository
33o4-mini-2025-04-16_mediumOpenAIclosedeffort medium75.0%not datedlechmazur/writing Github repository
34gemini-2.0-flash-thinking-exp-01-21Google DeepMind,Googlecloseddefault73.8%not datedlechmazur/writing Github repository
35claude-3-5-haiku-20241022Anthropiccloseddefault73.5%not datedlechmazur/writing Github repository
36grok-3-mini-beta_lowxAIclosedeffort low73.5%not datedlechmazur/writing Github repository
37zai-org/GLM-4.5Epoch: glm-4.5, Z.ai (Zhipu AI),Tsinghua Universityopendefault73.4%not datedlechmazur/writing Github repository
38qwen2.5-maxAlibabacloseddefault72.9%not datedlechmazur/writing Github repository
39gemini-2.0-flash-expGoogle DeepMind,Googlecloseddefault71.5%not datedlechmazur/writing Github repository
40o1-2024-12-17_mediumOpenAIclosedeffort medium70.2%not datedlechmazur/writing Github repository
41mistral-large-2407Mistral AIno page matchopendefault69.0%not datedlechmazur/writing Github repository
42gpt-4o-mini-2024-07-18OpenAIcloseddefault67.2%not datedlechmazur/writing Github repository
43o1-mini-2024-09-12_mediumOpenAIclosedeffort medium64.9%not datedlechmazur/writing Github repository
44grok-2-1212xAIcloseddefault63.6%not datedlechmazur/writing Github repository
45microsoft/phi-4Epoch: phi-4, Microsoft Researchopendefault62.6%not datedlechmazur/writing Github repository
46meta-llama/Llama-4-Maverick-17B-128E-InstructEpoch: Llama-4-Maverick-17B-128E-Instruct, Meta AIopendefault62.0%not datedlechmazur/writing Github repository
47o3-mini-2025-01-31_highOpenAIclosedeffort high61.7%not datedlechmazur/writing Github repository
48o3-mini-2025-01-31_mediumOpenAIclosedeffort medium61.5%not datedlechmazur/writing Github repository
49amazon.nova-pro-v1:0Amazoncloseddefault60.5%not datedlechmazur/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