Qwen 3 235B A22B FP8 (Thinking): benchmark results

Provider: Alibaba. Released 2025-04-29. Access: Open.

Unified ELO 1627 ± 1, rank #466 of 3078 rated models, from 29 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
BenchTable - STEM83.3Weighted Score (%)94.9
BenchTable67.6Total Score (%)83.4
LMGame-Bench Sokoban5.3Score81.2
BenchTable - Reasoning64.9Weighted Score (%)80.2
BenchTable - Tech70.8Weighted Score (%)71.9
AGI-Eval Community - Learning (English)92.67Accuracy (%)71
AGI-Eval Community - Learning87.55Accuracy (%)69.3
BenchTable - Utility61.7Weighted Score (%)65.6
AGI-Eval Community - Learning (Chinese)80.67Accuracy (%)63
LMGame-Bench Candy Crush437Score58.3
AGI-Eval Community - Subject Reasoning (Chinese)82.92Accuracy (%)52.9
AGI-Eval Community - General Reasoning96.22Accuracy (%)51.4

Interactive version: theaggregate.ai/model?slug=qwen-3-235b-a22b-fp8-thinking · How It Works · Data refreshed daily, snapshot 2026-09-19.