MiniMax M3.1 Flash vs Muse Wonder Full
AutomationBench-AA AutomationBench-AA: Agents orchestrate SaaS workflows through APIs; objectives receive credit only when the task has no guardrail violation
MiniMax M3.1 Flash72.5
Muse Wonder Full71.3
GDPval-AA (work tasks) 2.1 GDPval-AA (work tasks) 2.1: Real work tasks from many professions (reports, spreadsheets, analyses), graded against expert work
MiniMax M3.1 Flash47.9
Muse Wonder Full67.9
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
MiniMax M3.1 Flash25.8
Muse Wonder Full57.1
AA-Briefcase 1.1 · Elo AA-Briefcase 1.1 · Elo: Agents produce professional deliverables; blind pairwise comparisons produce the source Elo rating
MiniMax M3.1 Flash1,473.9
Muse Wonder Full1,729.3
AA-LCR (long context) 1.1 AA-LCR (long context) 1.1: Questions that need reading and connecting facts across very long documents (around 100K tokens)
MiniMax M3.1 Flash79.7
Muse Wonder Full83.7
FrontierCode 1.1 FrontierCode 1.1: Agents implement changes in real repositories, graded against maintainer-defined requirements
MiniMax M3.1 Flash38.6
Muse Wonder Full—
Harvey LAB-AA 1.1 Harvey LAB-AA 1.1: Legal agents work from case documents; every rubric criterion must pass and material hallucinations receive no credit
MiniMax M3.1 Flash—
Muse Wonder Full13.1
| Benchmark | MiniMax M3.1 Flash | Muse Wonder Full |
|---|---|---|
| AutomationBench-AA | 72.5 | 71.3 |
| GDPval-AA (work tasks) 2.1 | 47.9 | 67.9 |
| Terminal-Bench 4.0 | 25.8 | 57.1 |
| AA-Briefcase 1.1 · Elo | 1,473.9 | 1,729.3 |
| AA-LCR (long context) 1.1 | 79.7 | 83.7 |
| FrontierCode 1.1 | 38.6 | — |
| Harvey LAB-AA 1.1 | — | 13.1 |
Specs
| Spec | MiniMax M3.1 Flash | Muse Wonder Full |
|---|---|---|
| Input, per 1M | $0.15 | — |
| Output, per 1M | $0.60 | — |
| Cost per task? | $0.22 | — |
| Speed, tok/s | 105 | — |
| Context | 1M | 1M |
| Weights | Closed | Closed |
| Released | Oct 2026 | Oct 2026 |
| See images | ✓ | ✓ |
| Read PDFs | — | ? |
| Hear audio | — | ✓ |
| Watch video | ✓ | ✓ |
| Call tools | Partial | ✓ |
| Force a tool call | ? | ? |
| Return strict JSON | Partial | ? |
| Search the web | ? | ? |
| Run code | ? | ? |
| Use a computer | ? | ? |
What breaks when you switch
MiniMax M3.1 Flash → Muse Wonder Full
Only in Muse Code›
Muse Wonder Full requires a Muse Code subscription; max reasoning effort on the Model API Wonder model does not select Full.
Muse Wonder Full → MiniMax M3.1 Flash
Complete model response must be appended›
In multi-turn tools, preserve the complete assistant response, including reasoning_content and tool_calls, to maintain reasoning continuity.