Qwen3 235B A22B Instruct 2507
Qwen · released 2025-07-21 · apache-2.0 license
Qwen3-235B-A22B-Instruct-2507 is Alibaba's updated 235B (22B active) open-weight non-reasoning MoE (July 2025, Apache-2.0): MMLU-Pro 83.0, GPQA 77.5, AIME25 70.3. Artificial Analysis index 25.
Key specs
| Type | Local open-weight |
|---|---|
| Parameters | 235.09B total · MoE, — active |
| Architecture | qwen3_moe |
| Context window | 262K tokens |
| Knowledge cutoff | 2025-06-30 |
| Modalities | text |
| Recommended backends | — |
| Minimum viable rig | — |
Benchmark scores
| GPQA Diamond | 77.5% |
|---|---|
| SWE-bench Verified | — |
| AIME | 70.3% |
| MMLU-Pro | 83% |
| BFCL v3 (tool use) | — |
| Composite score | 7.9 |
| Community rating | No reviews yet |
VRAM & disk per quantization
| Quant | VRAM | Disk | RAM | Context |
|---|---|---|---|---|
| Q4_K_M | 137.9 GB | 136.4 GB | — | 262K |
Strengths & weaknesses
Strengths: Strong knowledge and instruction-following for a non-reasoning model; Very long context with high RULER (92.5); Apache-2.0
Weaknesses: Non-reasoning mode lags dedicated reasoners on hard math/coding; Verbose per Artificial Analysis