QuarkMedBench (1,000-Word Limit) - Important: leaderboard
Metric: Important score rate (%): share of the Important rubric points (logical completeness and explanatory depth) that the answer earns, on a stratified sample of 2,000 real Chinese online medical and health queries (single- and multi-turn), answers scored against query-specific rubrics (Essential, Important, Highlight and Pitfall items) by a Qwen3-235B judge, with a strict limit of 1,000 words per answer; higher is better. Source: arxiv.org. Saturation forecast: Around 2029. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.2 | 73.33 | #105 |
| 2 | Qwen 3 235B A22B (Non-reasoning) | 69.62 | #304 (Qwen 3 235B A22B) |
| 3 | GPT-5.1 | 69.21 | #131 |
| 4 | Qwen 3 32B (Non-reasoning) | 67.38 | #424 (Qwen 3 32B) |
| 5 | Qwen 3 30B A3B 2507 (Thinking) | 66.57 | #366 (Qwen 3 30B A3B 2507) |
| 6 | Gemini 3 Pro | 65.81 | #77 |
| 7 | DeepSeek V3.2 | 63.94 | #198 |
| 8 | Qwen 3 30B A3B (Non-reasoning) | 63.85 | #488 (Qwen 3 30B A3B) |
| 9 | DeepSeek R1 0528 | 63.4 | #217 |
| 10 | Gemini 2.5 Pro | 59.18 | #145 |
| 11 | Qwen 3 8B (Non-reasoning) | 57.87 | #667 (Qwen 3 8B) |
| 12 | Seed-1.6 | 55.88 | #257 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=quarkmedbench-1-000-word-limit-important · How It Works · Data refreshed daily, snapshot 2026-10-11.