Qwen 3 235B A22B 2507 Instruct: benchmark results
Dedicated direct-answer checkpoint from the July Qwen3 update. Unlike the original hybrid release, it does not emit thinking blocks; the update targets instruction following and tool use. Provider: Alibaba. Released 2025-07-21. Access: Open.
Aggregate Rating 1656 ± 5, rank #300 of 1656 rated models, from 1051 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| BhashaBench - Ayur - Hindi | 54.78 | Accuracy (%) | 100 |
| BhashaBench - Legal | 76.68 | Accuracy (%) | 100 |
| BhashaBench - Legal - English | 80.15 | Accuracy (%) | 100 |
| CVDP-ECov (Agentic) | 69.9 | Coverage pass rate (%): share of the 83 hardware repositorie | 100 |
| Drift-Bench (Service-Oriented, No Clarification) - Expression | 75.33 | Success rate (%; 150 StableToolBench G1 API tasks with an ex | 100 |
| Drift-Bench (Service-Oriented, No Clarification) - Parameter | 67 | Success rate (%; 150 StableToolBench G1 API tasks with a par | 100 |
| Drift-Bench (State-Oriented, No Clarification) - Intention | 56.16 | Success rate (%; 200 AgentBench operating-system and databas | 100 |
| Drift-Bench (State-Oriented, With Clarification) - Expression | 70.17 | Success rate (%; 200 AgentBench operating-system and databas | 100 |
| Drift-Bench (State-Oriented, With Clarification) - Intention | 66.67 | Success rate (%; 200 AgentBench operating-system and databas | 100 |
| Drift-Bench (State-Oriented, With Clarification) - Premise | 68.67 | Success rate (%; 200 AgentBench operating-system and databas | 100 |
| Galileo Agent - Insurance Accuracy | 74 | Accuracy (%) | 100 |
| Kaggle ITBench - FinOps | 73.33 | Score (%) | 100 |
Interactive version: theaggregate.ai/model?slug=qwen-3-235b-a22b-2507-instruct · How It Works · Data refreshed daily, snapshot 2026-10-20.