Qwen 3 235B A22B 2507: benchmark results
The July 2025 update of Alibaba's open Qwen 3 235B MoE (22B active). Provider: Alibaba. Released 2025-07-01. Access: Open.
Aggregate Rating 1675 ± 5, rank #260 of 1656 rated models, from 932 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AgentDrive - Physics | 67.5 | Accuracy (%) | 100 |
| CFEval | 2134 | Score (points) | 100 |
| DeepSynth-Eval (Agentic Multi-Turn) | 35.5 | Overall checklist score (%; plan, per-section selective full | 100 |
| EsoBench | 29.4 | Score | 100 |
| FinMCP-Bench - Multi-Tool | 69.42 | Tool F1 (%): the harmonic mean of tool precision and tool re | 100 |
| GeoChallenge (Text Only) | 75.55 | Exact match in percent on all 908 problems on GeoChallenge-s | 100 |
| GeoChallenge (Text Only) - Easy | 81.09 | Exact match in percent on the Easy split on GeoChallenge-sma | 100 |
| GeoChallenge (Text Only) - Hard | 66.3 | Exact match in percent on the Hard split on GeoChallenge-sma | 100 |
| GeoChallenge (Text Only) - Medium | 75.95 | Exact match in percent on the Medium split on GeoChallenge-s | 100 |
| KAware - External Function Pass Rate | 98.04 | The end-to-end pass rate (%): share of tasks whose trajector | 100 |
| LLM Stats (WritingBench) | 88.3 | Score (%) | 100 |
| LiveFMBench - Pre2025 | 34.92 | Pass@1 (%) after the faithfulness check on LiveFMBench-Pre20 | 100 |
Interactive version: theaggregate.ai/model?slug=qwen-3-235b-a22b-2507 · How It Works · Data refreshed daily, snapshot 2026-10-20.