Qwen3 30B-A3B Instruct 2507
| Released | 2025-07-28 |
|---|---|
| Price per 1M | $0.10 in · $0.30 out |
| Overall score | 42.2 · rank 182 |
| Context window | 262,144 tokens |
| Provider | Alibaba |
| Weights | Open weights (Apache 2.0) |
Scores
| Dimension | Score | Rank | Benchmarks |
|---|---|---|---|
| Overall: Balanced | 42.2 | 182 | 4 |
| Overall: Coding | 39.8 | 189 | 2 |
| Overall: Reasoning | 43.0 | 173 | 2 |
| Overall: Agentic work | 39.8 | 187 | 2 |
| Agents | 39.8 | 156 | 1 |
| Reasoning | 39.7 | 197 | 1 |
| Math | 47.2 | 136 | 1 |
| General | 46.0 | 138 | 1 |
| Value vs price | +0.1 | 119 |
Evidence
Agents 39.8 ±3.6
Reasoning 39.7 ±4.6
| Benchmark | Result | Source | Date |
|---|---|---|---|
| GPQA Diamond | 55.6% | Epoch AI | 2026-08-30 |
Math 47.2 ±3.6
| Benchmark | Result | Source | Date |
|---|---|---|---|
| OTIS Mock AIME 2024-2025 | 62.2% | Epoch AI | 2026-08-30 |
General 46.0 ±4.1
| Benchmark | Result | Source | Date |
|---|---|---|---|
| LMArena Text | 1,382 | LMArena | 2026-11-09 |
Results not used in the score
| Benchmark | Result | Source | Why not used |
|---|---|---|---|
| BFCL v4 | 36.7% | BFCL | A stronger configuration result was used |
Names matched to this model
qwen3-30b-a3b-instruct-2507