Mistral Small 3.2 24B vs Mistral Small 4
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
Mistral Small 3.2 24B4.3
Mistral Small 49.9
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
Mistral Small 3.2 24B0
Mistral Small 40
AutomationBench-AA AutomationBench-AA: Agents orchestrate SaaS workflows through APIs; objectives receive credit only when the task has no guardrail violation
Mistral Small 3.2 24B0.3
Mistral Small 41
Arena Text · creative writing Arena Text · creative writing: Human pairwise preferences in the stated text vote pool
Mistral Small 3.2 24B1,324
Mistral Small 4—
MMMU-Pro (vision) MMMU-Pro (vision): College-level questions that require understanding images: charts, diagrams, photos
Mistral Small 3.2 24B48
Mistral Small 456.8
| Benchmark | Mistral Small 3.2 24B | Mistral Small 4 |
|---|---|---|
| Humanity's Last Exam | 4.3 | 9.9 |
| Terminal-Bench 4.0 | 0 | 0 |
| AutomationBench-AA | 0.3 | 1 |
| Arena Text · creative writing | 1,324 | — |
| MMMU-Pro (vision) | 48 | 56.8 |
Specs
| Spec | Mistral Small 3.2 24B | Mistral Small 4 |
|---|---|---|
| Input, per 1M | $0.10 | $0.15 |
| Output, per 1M | $0.30 | $0.60 |
| Cost per task? | $0.11 | $0.015 |
| Speed, tok/s | 30 | 212 |
| Context | 33K | 262K |
| Weights | Open | Open |
| Released | Jun 2025 | Mar 2026 |
| Memory (4-bit) | 14 GB | 72 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
Mistral Small 3.2 24B → Mistral Small 4
Memory (4-bit): 14 GB → 72 GB
Mistral Small 4 → Mistral Small 3.2 24B