Qwen

Comparing 10 models from Qwen.

Model Total ▼ Released Context CoTTooling Creative Writing Language Utility Reasoning Text Editing Rule Following Hallucination
Qwen 3.5 397B A17B91.73%Feb 15, 26128k99.77%86.93%95.01%97.50%95.06%98.05%79.39%82.10%
Qwen 3.5 122B91.53%Feb 25, 26262k100.00%83.02%95.01%96.36%94.93%96.31%80.00%86.58%
Qwen 3.5 27B90.85%Feb 25, 26262k99.00%82.54%95.52%95.67%92.73%98.69%76.04%86.58%
Qwen 3.5 35B88.00%Feb 25, 26262k93.98%83.51%91.95%96.42%94.88%94.95%67.42%80.87%
Qwen 3.5 Flash86.38%Feb 25, 261m87.87%83.81%91.94%96.11%94.66%92.80%63.19%80.63%
Qwen 3.5 9B86.05%Mar 10, 26262k97.00%84.35%88.18%94.02%92.93%85.35%60.98%85.58%
Qwen 3.5 Plus (2026-02-15)85.96%Feb 15, 261m99.74%77.07%95.10%86.65%93.45%98.10%64.21%73.35%
Qwen 3 32B82.21%Apr 28, 2541k97.43%81.30%84.61%81.66%86.35%89.95%46.83%89.56%
Qwen3 235B A22B Instruct 250780.10%Jul 21, 25262.1k99.23%84.81%60.83%83.15%85.82%91.75%65.42%69.82%
Qwen 2.5 72B75.46%Sep 19, 24131.1k99.38%75.16%68.95%76.43%83.43%89.18%31.55%79.56%
Model Performance
Cost vs Performance

Compares total benchmark cost against overall score for Qwen models. Quadrant lines are drawn at the median values.

Cost Breakdown

Total benchmark cost per model, broken down by input, reasoning, and output tokens. Toggle between USD and token views.