Qwen 3 235B A22B FP8: benchmark results

Qwen 3 235B A22B evaluated in FP8 quantization. Provider: Alibaba. Released 2025-04-28. Access: Open.

Unified ELO 1609 ± 1, rank #239 of 1392 rated models, from 12 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
Kluster Hallucination Detection - RAG Method 3 Resistance93.37Resistance (100 - Hallucination Rate %)70
ReasonScape M12X473.79ReasonScore (hard)65.8
Kluster Hallucination Detection - RAG Method 1 Resistance93.2Resistance (100 - Hallucination Rate %)63.6
StickToYourRole71.84Cardinal Score62.9
BenchTable55.5Total Score (%)62.8
LMGame-Bench Candy Crush363.3Score50
LMGame-Bench 20482144Score45.8
LMGame-Bench Sokoban2.3Score45.8
Kluster Hallucination Detection - Non-RAG Resistance97.91Resistance (100 - Hallucination Rate %)43.8
Kluster Hallucination Detection - RAG Method 2 Resistance94.96Resistance (100 - Hallucination Rate %)36.4
Kluster Hallucination Detection - RAG Resistance93.84Resistance (100 - Hallucination Rate %)31.2
LMGame-Bench Tetris11.7Score10.4

Interactive version: theaggregate.ai/model?slug=qwen-3-235b-a22b-fp8 · How It Works · Data refreshed daily, snapshot 2026-09-05.