MiMo-V2.5-Pro vs Qwen3 8
MiMo-V2.5-Pro 
Qwen3 8 Verdict
- Qwen3 8 (max) leads on quality: 66.5 vs 59.2.
- MiMo-V2.5-Pro is stronger in instruction following.
- Qwen3 8 (max) is stronger in agents & tools, coding, composite, human preference, knowledge, long context, maths, reasoning, multimodal.
- MiMo-V2.5-Pro is 4.1× cheaper ($0.653 vs $2.67 per 1M blended).
- MiMo-V2.5-Pro streams 1.4× faster (53 vs 39 tokens per second).
| Metric | MiMo-V2.5-Pro | Qwen3 8 (max) |
|---|---|---|
| BenchLeader Index | 59.2 | 66.5 |
| Agents & tools score | 45.9 | 68.3 |
| Coding score | 58.3 | 66.5 |
| Composite score | 62.5 | 74.3 |
| Human preference score | 61.5 | 68.2 |
| Instruction following score | 77.3 | – |
| Knowledge score | 57.6 | 63.5 |
| Long context score | 66.3 | 66.6 |
| Maths score | 59.0 | 67.0 |
| Reasoning score | 55.6 | 76.3 |
| Multimodal score | – | 67.4 |
| Blended price $/M | $0.653 | $2.67 |
| Output speed | 53 tok/s | 39 tok/s |
| Time to first answer | 42.2 s | 55.1 s |
| Context window | 1.0M | 1M |
| SciCode | 50.2% | 52.9% |
| ProofBench | 22.0% | 58.0% |
| Epoch Capabilities Index | – | 156.6 |
| LMArena Text | 1467 | 1481 |
| LMArena Hard Prompts | 1496 | 1503 |
| LMArena Coding | 1521 | 1522 |
| LMArena WebDev | 1475 | 1671 |
| LMArena Vision | – | 1315 |
| LMArena Agent | -5.7 | 3.3 |
| LiveBench | – | 78.5% |
| LiveBench Reasoning | – | 88.2% |
| LiveBench Coding | – | 72.9% |
| LiveBench Agentic Coding | – | 64.7% |
| LiveBench Mathematics | – | 91.3% |
| LiveBench Data Analysis | – | 78.4% |
| LiveBench Language | – | 79.7% |
| LiveBench Instruction Following | – | 74.1% |
| AA Intelligence Index | 26.4 | 45.4 |
| IFBench | 79.9% | – |
| AA-LCR | 79.7% | 80.3% |
| MMMU-Pro | – | 82.8% |
| AA-Omniscience | 3.3 | 12.0 |
| Terminal-Bench Hard | 43.2% | – |
| GPQA Diamond (AA) | 86.6% | 92.8% |
| Humanity's Last Exam (AA) | 35.7% | 43.1% |
| SciCode (AA) | 50.6% | 53.2% |
| τ²-Bench Telecom (AA) | 94.2% | – |
| LiveCodeBench | 81.3% | 87.8% |
| MMLU-Pro | 84.6% | 88.6% |
| IOI | – | 68.9% |
| LegalBench | 77.1% | 83.6% |
| CorpFin | 61.4% | 65.8% |
| TaxEval | 73.8% | 75.5% |
| Terminal-Bench 2.1 (Vals) | 57.3% | 67.4% |
| SWE-bench (Vals) | 74.0% | 85.6% |
| GPQA Diamond (Vals) | 82.6% | 93.7% |
| Vals Index | 41.0 | 51.8 |
| EQ-Bench 4 | 1208 | – |
| CritPt | 4.0% | 20.0% |
| GDPval (AA) | 34.3% | 58.2% |
| τ²-Bench Banking (AA) | 9.9% | 51.3% |
| ITBench SRE (AA) | 38.2% | – |
| Analyst Agent (AA) | 20.0% | – |
| APEX-Agents (AA) | 2.4% | – |
| Code Migration | 21.6% | 24.0% |
| Excel Modeling Benchmark | 55.2% | 60.1% |
| Finance Agent v2 | 41.5% | 50.6% |
| Harvey's Legal Agent Benchmark | 2.1% | 10.4% |
| Legal Research Bench | 15.9% | 47.6% |
| MedCode | 32.5% | 40.7% |
| MedScribe | 83.7% | 85.0% |
| MMMU-Pro (Vals) | – | 88.0% |
| MortgageTax | – | 64.0% |
| MysteryMechanism | – | 23.9% |
| ProgramBench | – | 0.0% |
| Public Benefits Bench | – | 67.1% |
| SAGE | – | 51.3% |
| SkillsBench | – | 42.0% |
| Tax Agent Bench | – | 66.0% |
| Terminal-Bench 4.0 (Vals) | – | 24.8% |
| Terminal-Bench Science | – | 1.4% |
| Vals Multimodal Index | – | 65.4% |
| Vibe Code Bench 1-100 | – | 12.8% |
| Vibe Code Bench v1.1 | 34.1% | 64.7% |
| LMArena Maths | 1476 | 1498 |
| LMArena Creative Writing | 1435 | 1472 |
| LMArena Instruction Following | 1470 | 1477 |
| LMArena Multi-turn | 1478 | 1495 |
| LMArena Longer Queries | 1487 | 1492 |
| LMCA | 29.5% | – |
| DTBench | 84.5% | – |
| ALE-Bench | 899.8 | – |
Data as of 2026-09-19. Best configuration of each model; every score links to its source on the model pages.
MiMo-V2.5-Pro vs Qwen3 8: questions
- Is MiMo-V2.5-Pro better than Qwen3 8?
- Qwen3 8 (max) leads on quality: 66.5 vs 59.2. The BenchLeader Index combines every independent quality benchmark; Qwen3 8 (max) is ahead overall as of 2026-09-19, but check the category scores for your use.
- Is MiMo-V2.5-Pro better than Qwen3 8 for coding?
- Qwen3 8 scores higher in coding (67 vs 58 on the category index, where 50 is average).
- Is MiMo-V2.5-Pro better than Qwen3 8 for agentic tasks?
- Qwen3 8 scores higher in agentic tasks (68 vs 46 on the category index, where 50 is average).
- Which is cheaper, MiMo-V2.5-Pro or Qwen3 8?
- MiMo-V2.5-Pro is cheaper: $0.653 against $2.67 per million tokens, blended at three input tokens per output token.
- Which is faster, MiMo-V2.5-Pro or Qwen3 8?
- MiMo-V2.5-Pro streams faster: 53 against 39 output tokens per second.
- Which has the larger context window?
- MiMo-V2.5-Pro accepts more context: 1.0M against 1M tokens.