MiniMax M2.1 vs MiniMax M3
Humanity's Last Exam Humanity's Last Exam: Very hard expert questions across many fields, written to stump today’s models
MiniMax M2.123.2
MiniMax M339
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
MiniMax M2.1—
MiniMax M32
AutomationBench-AA AutomationBench-AA: Agents orchestrate SaaS workflows through APIs; objectives receive credit only when the task has no guardrail violation
MiniMax M2.1—
MiniMax M321.3
Arena Text · creative writing Arena Text · creative writing: Human pairwise preferences in the stated text vote pool
MiniMax M2.1—
MiniMax M31,405
MMMU-Pro (vision) MMMU-Pro (vision): College-level questions that require understanding images: charts, diagrams, photos
MiniMax M2.1—
MiniMax M378.6
| Benchmark | MiniMax M2.1 | MiniMax M3 |
|---|---|---|
| Humanity's Last Exam | 23.2 | 39 |
| Terminal-Bench 4.0 | — | 2 |
| AutomationBench-AA | — | 21.3 |
| Arena Text · creative writing | — | 1,405 |
| MMMU-Pro (vision) | — | 78.6 |
Specs
| Spec | MiniMax M2.1 | MiniMax M3 |
|---|---|---|
| Input, per 1M | $0.30 | $0.30 |
| Output, per 1M | $1.20 | $1.20 |
| Cost per task? | — | $0.51 |
| Speed, tok/s | 72 | 157 |
| Context | 205K | 524K |
| Weights | Open | Open |
| Released | Dec 2025 | Jun 2026 |
| Memory (4-bit) | 137 GB | 256 GB |
| Licence | modified-mit | minimax-community |
| Commercial use | Restricted | Restricted |
| See images | — | ✓ |
| Read PDFs | — | Partial |
| Hear audio | — | — |
| Watch video | — | ✓ |
| Call tools | ✓ | ✓ |
| Force a tool call | Partial | Partial |
| Return strict JSON | Partial | — |
| Search the web | ? | ? |
| Run code | ? | ? |
| Use a computer | ? | ? |
What breaks when you switch
MiniMax M2.1 → MiniMax M3
Memory (4-bit): 137 GB → 256 GB
No strict JSON
Deployment status experimental · Self-hosted›
The documented MiniMax M3 baseline uses an experimental SGLang development image; pin its model revision and image digest and validate your workload.
MiniMax M3 → MiniMax M2.1