Qwen3
Qwen3-235B-A22B-Instruct-2507
Alibaba (Qwen team) · open-permissive apache-2.0
| Parameters (B) | 235.094 |
|---|---|
| Architecture | moe |
| Layers / KV heads / head dim | 94 / 4 / 128 |
| Context window | 262144 |
| Modalities | in: text; out: text |
| GGUF | available |
| Commercial use | yes |
Minimum hardware (estimate)
At Q4_K_M and 8k context, total working set ≈ 144.9 GiB. On a machine with 64 GiB system RAM and no discrete GPU: Won't run. Limiting factor: No GPU and system RAM insufficient for weights + KV.
Scores
| Benchmark | Value | Uncertainty | Run by | Source |
|---|---|---|---|---|
| GPQA Diamond | 80.0% | ±0.0260 | third-party | epoch-ai |
| MATH Level 5 | 68.9% | ±0.0095 | third-party | epoch-ai |
| OTIS Mock AIME 2024-2025 | 86.7% | ±0.0512 | third-party | epoch-ai |
| FrontierMath | 8.5% | ±0.0166 | third-party | epoch-ai |
| FrontierMath Tier 4 | 0.0% | ±0.0000 | third-party | epoch-ai |
| SimpleQA Verified | 50.1% | ±0.0158 | third-party | epoch-ai |
| Aider polyglot | 59.6% | unknown | third-party | epoch-ai |
Provenance
- identity, license, architecture, parameter count: https://huggingface.co/Qwen/Qwen3-235B-A22B-Instruct-2507 (retrieved 2026-08-10)
HF: Qwen/Qwen3-235B-A22B-Instruct-2507 · snapshot 2026-08-10