MiMo-V2.6-Pro vs Muse Wonder Full
As of 18 Oct 2026, Muse Wonder Full (max) is the better pick for most workloads: it leads on quality (68.6 vs 65.3 on the BenchLeader Index).
MiMo-V2.6-Pro 
Muse Wonder Full Key differences
- Muse Wonder Full (max) leads on quality: 68.6 vs 65.3.
- MiMo-V2.6-Pro is stronger in agents & tools, human preference, long context, maths, multimodal.
- Muse Wonder Full (max) is stronger in coding, composite, knowledge, reasoning.
Choose MiMo-V2.6-Pro if
- long context is the priority (68 vs 67 on the category score)
- you want to run it yourself: its weights are open
Choose Muse Wonder Full (max) if
- reasoning is the priority (95 vs 79 on the category score)
- knowledge is the priority (71 vs 65 on the category score)
- coding is the priority (64 vs 62 on the category score)
Category by category
Category scores on the BenchLeader scale: 50 is the average evaluated model, 15 points is one standard deviation. A gap under one point is called level.
| Category | MiMo-V2.6-Pro | Muse Wonder Full (max) | Leader |
|---|---|---|---|
| Reasoning | 79.0 | 95.0 | Muse Wonder Full |
| Coding | 61.6 | 64.0 | Muse Wonder Full |
| Agents & tools | 60.2 | – | not comparable |
| Maths | 66.3 | – | not comparable |
| Knowledge | 65.2 | 71.3 | Muse Wonder Full |
| Human preference | 67.4 | – | not comparable |
| Multimodal | 60.1 | – | not comparable |
| Long context | 68.3 | 67.0 | MiMo-V2.6-Pro |
| Composite | 81.5 | 92.2 | Muse Wonder Full |
Price, speed and latency
List prices blended at three input tokens per output token. Per-task prices use the token counts on the methodology page; suite cost is what Artificial Analysis paid to run its whole Intelligence Index.
| Measure | MiMo-V2.6-Pro | Muse Wonder Full (max) |
|---|---|---|
| Blended price per 1M tokens | $0.548 | – |
| Input / output per 1M | $0.440 / $0.870 | – / – |
| Price per task: chat reply | $0.0004 | – |
| Price per task: summarise a 30-page report | $0.0058 | – |
| Price per task: code edit | $0.0039 | – |
| Price per task: agentic coding session | $0.030 | – |
| Price per task: structured extraction | $0.0011 | – |
| Cost to run a full benchmark suite | $0.133 | – |
| Output speed | 48 tok/s | – |
| Time to first answer | 45.8 s | – |
| Time to a complete answer | 56.2 s | – |
Every benchmark
| Metric | MiMo-V2.6-Pro | Muse Wonder Full (max) |
|---|---|---|
| BenchLeader Index | 65.3 | 68.6 |
| Agents & tools score | 60.2 | – |
| Coding score | 61.6 | 64.0 |
| Composite score | 81.5 | 92.2 |
| Human preference score | 67.4 | – |
| Knowledge score | 65.2 | 71.3 |
| Long context score | 68.3 | 67.0 |
| Maths score | 66.3 | – |
| Multimodal score | 60.1 | – |
| Reasoning score | 79.0 | 95.0 |
| Blended price $/M | $0.548 | – |
| Output speed | 48 tok/s | – |
| Time to first answer | 45.8 s | – |
| Context window | 1.0M | 1M |
| SciCode | 60.9% | – |
| APEX-Agents | 59.5% | – |
| ProofBench | 70.0% | – |
| Arena Text | 1479 | – |
| Arena Hard Prompts | 1504 | – |
| Arena Coding | 1538 | – |
| Arena WebDev | 1625 | – |
| Arena Vision | 1264 | – |
| Arena Agent | 4 | – |
| AA Intelligence Index v4.3.2 | 46.3 | 55.8 |
| AA-LCR | 86.3% | 83.7% |
| AA-Omniscience | 8.4 | 30.1 |
| Humanity's Last Exam (AA) | 49.4% | 53.6% |
| SciCode (AA) | 60.9% | 62.6% |
| IOI | 39.3% | 79.8% |
| LegalBench | – | 87.4% |
| Terminal-Bench 2.1 (Vals) | 67.8% | – |
| Vals Index | 55.2 | 64.6 |
| CritPt | 26.6% | 32.9% |
| GDPval-AA v2.1 | 59.2% | 67.8% |
| Analyst Agent (AA) | 37.5% | – |
| BioMysteryBench | – | 73.3% |
| Code Migration | 43.0% | 61.0% |
| CyberBench | 72.9% | 73.6% |
| Excel Modeling Benchmark | 62.9% | 71.9% |
| Finance Agent v2 | 57.3% | 59.1% |
| Harvey's Legal Agent Benchmark | 10.8% | 21.7% |
| Legal Research Bench | 47.1% | 53.4% |
| MedCode | 45.0% | 49.9% |
| MedScribe | 88.3% | 89.3% |
| MysteryMechanism | 15.3% | 35.1% |
| ProgramBench | 0.5% | 0.5% |
| Public Benefits Bench | 68.9% | 73.8% |
| SAGE | 45.0% | – |
| SREBench | 3.0% | 11.8% |
| Tax Agent Bench | 32.4% | 43.0% |
| Terminal-Bench 4.0 (Vals) | 31.3% | 58.1% |
| Terminal-Bench Science | 2.9% | 37.1% |
| Vibe Code Bench 1-100 | – | 34.0% |
| Vibe Code Bench v1.1 | 85.2% | 86.0% |
| Arena Maths | 1488 | – |
| Arena Creative Writing | 1445 | – |
| Arena Instruction Following | 1471 | – |
| Arena Multi-turn | 1455 | – |
| Arena Longer Queries | 1494 | – |
| ALE-Bench | 1158.1 | – |
| Terminal-Bench 4.0 (AA) | 34.9% | 57.1% |
| AutomationBench | 58.6% | 71.3% |
| GDP.pdf | 19.2% | 32.6% |
| MLCR | 18.3% | – |
| Harvey LAB-AA v1.1 | – | 13.1% |
| AA-Omniscience: accuracy | 34.9% | 43.2% |
| AA-Omniscience: non-hallucination | 59.4% | 77.0% |
| AA-Briefcase v1.1 | 1515 | 1729 |
| FrontierCode 1.1 (AA) | 45.4% | – |
| Terminal-Bench-Science 0.1 (AA) | 5.7% | – |
| AA Cyber Index | 5606.6% | – |
| CWE-Bench (AA) | 63.3% | – |
| DeepSec-Bench (AA) | 26.2% | – |
| CyberGym end-to-end (AA) | 78.6% | – |
| AA-3D | 980 | – |
| AA Openness Index | 52.8 | – |
Data as of 2026-10-18. Best configuration of each model; every score links to its source on the model pages.
MiMo-V2.6-Pro vs Muse Wonder Full: questions
- Is MiMo-V2.6-Pro better than Muse Wonder Full?
- Muse Wonder Full (max) leads on quality: 68.6 vs 65.3. The BenchLeader Index combines every independent quality benchmark; Muse Wonder Full (max) is ahead overall as of 2026-10-18, but check the category scores for your use.
- Is MiMo-V2.6-Pro better than Muse Wonder Full for coding?
- Muse Wonder Full scores higher in coding (64 vs 62 on the category index, where 50 is average).
- Which has the larger context window?
- MiMo-V2.6-Pro accepts more context: 1.0M against 1M tokens.