legal benchmark
WritingBench Leaderboard
A comprehensive benchmark for evaluating large language models' generative writing capabilities across 6 core writing domains (Academic & Engineering, Finance & Business, Politics & Law, Literature & Art, Education, Advertising & Marketing) and 100 subdomains. Contains 1,239 queries with a query-dependent evaluation framework that dynamically generates 5 instance-specific assessment criteria for each writing task, using a fine-tuned critic model to score responses on style, format, and length dimensions.
Updated Aug 17, 2026
Models15
Model coverage15
MetricScore
EvidenceB
WritingBench Ranking
Higher score ranks better on this benchmark.
Sort by
Rank | Model | Score | Percentile | Participants | Evidence | Evaluated |
|---|
| Rank01 | ModelAC | Score88.3% | Percentile100.0% | Participants15 | EvidenceC | Evaluated |
| Rank02 | ModelAC | Score87.3% | Percentile92.9% | Participants15 | EvidenceC | Evaluated |
| Rank03 | ModelAC | Score86.7% | Percentile85.7% | Participants15 | EvidenceC | Evaluated |
| Rank04 | ModelAC | Score86.2% | Percentile78.6% | Participants15 | EvidenceC | Evaluated |
| Rank05 | ModelAC | Score85.5% | Percentile71.4% | Participants15 | EvidenceC | Evaluated |
| Rank06 | ModelAC | Score85.5% | Percentile64.3% | Participants15 | EvidenceC | Evaluated |
| Rank07 | ModelAC | Score85.2% | Percentile57.1% | Participants15 | EvidenceC | Evaluated |
| Rank08 | ModelAC | Score85.2% | Percentile50.0% | Participants15 | EvidenceC | Evaluated |
| Rank09 | ModelAC | Score84.6% | Percentile42.9% | Participants15 | EvidenceC | Evaluated |
| Rank10 | ModelAC | Score84.0% | Percentile35.7% | Participants15 | EvidenceC | Evaluated |
| Rank11 | ModelAC | Score83.1% | Percentile28.6% | Participants15 | EvidenceC | Evaluated |
| Rank12 | ModelAC | Score82.9% | Percentile21.4% | Participants15 | EvidenceC | Evaluated |
| Rank13 | ModelAC | Score82.6% | Percentile14.3% | Participants15 | EvidenceC | Evaluated |
| Rank14 | ModelAC | Score82.5% | Percentile7.1% | Participants15 | EvidenceC | Evaluated |
| Rank15 | ModelMA | Score73.8% | Percentile0.0% | Participants15 | EvidenceC | Evaluated |
WritingBench Highlights
The leading models and scores on this benchmark.
WritingBench Score Distribution
A closer view of the leading scores on this benchmark.
The Top AI Models for WritingBench
The first five results on this benchmark, with official price and output speed added where the model identity can be matched.
Ranking basisThis writingbench AI model leaderboard uses descending score in the benchmark's original unit. The leaderboard ranking keeps matched price and speed data separate from benchmark evidence.
- 01ACQwen3-235B-A22B-Thinking-2507Alibaba Cloud / Qwen Team
- Score
- 88.3%
Strengths
- Ranks #1 of 15 compared models
- 100th percentile on this benchmark
- C evidence result
Considerations
- This result measures WritingBench, not total model capability
- 02ACQwen3-Next-80B-A3B-InstructAlibaba Cloud / Qwen Team
- Score
- 87.3%
- Price
- $0.50 input / $2.0 output per 1M tokens
Strengths
- Ranks #2 of 15 compared models
- 93th percentile on this benchmark
- C evidence result
Considerations
- This result measures WritingBench, not total model capability
- 03ACQwen3 VL 235B A22B ThinkingAlibaba Cloud / Qwen Team
- Score
- 86.7%
Strengths
- Ranks #3 of 15 compared models
- 86th percentile on this benchmark
- C evidence result
Considerations
- This result measures WritingBench, not total model capability
- 04ACQwen3 VL 32B ThinkingAlibaba Cloud / Qwen Team
- Score
- 86.2%
Strengths
- Ranks #4 of 15 compared models
- 79th percentile on this benchmark
- C evidence result
Considerations
- This result measures WritingBench, not total model capability
- 05ACQwen3 VL 235B A22B InstructAlibaba Cloud / Qwen Team
- Score
- 85.5%
Strengths
- Ranks #5 of 15 compared models
- 71th percentile on this benchmark
- C evidence result
Considerations
- This result measures WritingBench, not total model capability
Selection summary
Best AI Models for WritingBench
Qwen3-235B-A22B-Thinking-2507 currently leads WritingBench with 88.3%. It is the top model on this specific benchmark, while the best LLM for the broader task should also be checked against other benchmarks, price and runtime.
Use this leaderboard with the supporting benchmark results and coverage details above. A leaderboard position summarizes the selected ranking signal; it does not replace workload-specific testing.
What is WritingBench?
What WritingBench measures and how its scores work.
A comprehensive benchmark for evaluating large language models' generative writing capabilities across 6 core writing domains (Academic & Engineering, Finance & Business, Politics & Law, Literature & Art, Education, Advertising & Marketing) and 100 subdomains. Contains 1,239 queries with a query-dependent evaluation framework that dynamically generates 5 instance-specific assessment criteria for each writing task, using a fine-tuned critic model to score responses on style, format, and length dimensions.
Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.
- Family
- WritingBench
- Modality
- text
- Primary category
- legal
- Score direction
- higher
- LLMBoard eligible
- Yes
- Evaluation key
- writingbench|llm-stats-current
Benchmark scores retain their original unit. Overall score eligibility is shown separately.
FAQ
Common questions about WritingBench.
Which model scores highest on WritingBench?
Qwen3-235B-A22B-Thinking-2507 is currently ranked first with 88.3%.
What does WritingBench measure?
A comprehensive benchmark for evaluating large language models' generative writing capabilities across 6 core writing domains (Academic & Engineering, Finance & Business, Politics & Law, Literature & Art, Education, Advertising & Marketing) and 100 subdomains. Contains 1,239 queries with a query-dependent evaluation framework that dynamically generates 5 instance-specific assessment criteria for each writing task, using a fine-tuned critic model to score responses on style, format, and length dimensions.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
15 model results are currently shown.
Does this benchmark affect the overall score?
Yes. This benchmark can contribute to the current LLMBoard capability score.