Qwen3 VL 30B A3B Instruct vs Qwen3.6 35B A3B
Different efforts: non-thinking · thinking
Humanity's Last Exam Humanity's Last Exam: Very hard expert questions across many fields, written to stump today’s models
Qwen3 VL 30B A3B Instruct6.3
Qwen3.6 35B A3B22.2
Terminal-Bench 4.0 Terminal-Bench 4.0 · source setting 2: Agents use a terminal to complete executable coding, system administration and research tasks in the 4.0 suite
Qwen3 VL 30B A3B Instruct—
Qwen3.6 35B A3B0
AutomationBench-AA AutomationBench-AA: Agents orchestrate SaaS workflows through APIs; objectives receive credit only when the task has no guardrail violation
Qwen3 VL 30B A3B Instruct—
Qwen3.6 35B A3B5.2
MMMU-Pro (vision) MMMU-Pro (vision): College-level questions that require understanding images: charts, diagrams, photos
Qwen3 VL 30B A3B Instruct62.1
Qwen3.6 35B A3B75
| Benchmark | Qwen3 VL 30B A3B Instruct | Qwen3.6 35B A3B |
|---|---|---|
| Humanity's Last Exam | 6.3 | 22.2 |
| Terminal-Bench 4.0 | — | 0 |
| AutomationBench-AA | — | 5.2 |
| MMMU-Pro (vision) | 62.1 | 75 |
Specs
| Spec | Qwen3 VL 30B A3B Instruct | Qwen3.6 35B A3B |
|---|---|---|
| Input, per 1M | $0.15 | $0.25 |
| Output, per 1M | $0.50 | $1.49 |
| Cost per task? | — | $0.48 |
| Speed, tok/s | 28 | 138 |
| Context | 256K | 262K |
| Weights | Open | Open |
| Released | Oct 2025 | Apr 2026 |
| Memory (4-bit) | 19 GB | 22 GB |
| Licence | Apache-2.0 | Apache-2.0 |
| Commercial use | Allowed | Allowed |
| See images | ✓ | ✓ |
| Read PDFs | — | — |
| Hear audio | — | — |
| Watch video | — | ✓ |
| Call tools | ✓ | ✓ |
| Force a tool call | Partial | Partial |
| Return strict JSON | ✓ | ✓ |
| Search the web | ? | ? |
| Run code | ? | ? |
| Use a computer | ? | ? |
What breaks when you switch
Qwen3 VL 30B A3B Instruct → Qwen3.6 35B A3B
Memory (4-bit): 19 GB → 22 GB
Qwen3.6 35B A3B → Qwen3 VL 30B A3B Instruct