MiMo v2.5 Pro vs MiMo v2.6 Pro
| Benchmark | MiMo v2.5 Pro | MiMo v2.6 Pro |
|---|---|---|
| Humanity's Last Exam | 35.7 | 49.4 |
| Arena Text · overall | 1,467 | 1,479 |
| Arena Text · Chinese | 1,506.1 | 1,530.5 |
| Arena Text · English | 1,476.7 | 1,504.4 |
| Arena Text · instruction following | 1,475.2 | 1,490.6 |
| Arena Text · non-English | 1,448.2 | 1,473.7 |
| Arena Text · Russian | 1,449.2 | 1,478.8 |
| AA Omniscience Index | 3.3 | 8.4 |
| CritPt (physics) | 4 | 26.6 |
| AA Intelligence Index 4.3.2 | 26 | 46.3 |
| AA Omniscience · Accuracy | 22.4 | 34.9 |
| AA Omniscience · Hallucination Rate | 24.7 | 40.6 |
| AA-LCR (long context) 1.1 | 79.7 | 86.3 |
| GDP.pdf · AA all-pass | 4 | 19.2 |
| MLCR-AA | 5.6 | 12.8 |
| Vals MedCode | 32.5 | 45 |
| Vals MedScribe | 83.7 | 88.3 |
| Vals Legal Research Bench | 15.9 | 47.1 |
| Arena Text · French | 1,487.1 | — |
| Arena Text · German | 1,459.2 | — |
| Arena Text · Japanese | 1,409.9 | — |
| Arena Text · Korean | 1,435.9 | — |
| Arena Text · Polish | 1,463.1 | — |
| Arena Text · Spanish | 1,467.8 | — |
| IFBench (instructions) | 79.9 | — |
| Vals LegalBench | 77.1 | — |
Specs
| Spec | MiMo v2.5 Pro | MiMo v2.6 Pro |
|---|---|---|
| Input, per 1M | $0.43 | $0.43 |
| Output, per 1M | $0.87 | $0.87 |
| Cost per task? | $0.15 | $0.13 |
| Speed, tok/s | 52 | 55 |
| Context | 1M | 1M |
| Weights | Open | Open |
| Released | Apr 2026 | Sep 2026 |
| Memory (4-bit) | 614 GB | 615 GB |
| Licence | MIT | MIT |
| 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
MiMo v2.5 Pro → MiMo v2.6 Pro
Memory (4-bit): 614 GB → 615 GB
Pass back reasoning_content›
In deep-thinking multi-turn tool conversations, return the full reasoning_content in later rounds with tool calls or receive HTTP 400.
Forced recommended defaults›
Deep thinking overrides supplied temperature and top_p with 1.0 and 0.95.
Avoid answer truncation›
max_completion_tokens includes thinking and the final answer; reserve enough tokens to avoid truncating the answer.
MOPD checkpoints address tool-call repetition›
Use the MOPD checkpoints for the tool-call repetition fix; the API has served the updated models since September 25 at 06:00 UTC+8 under unchanged names.
MiMo v2.6 Pro → MiMo v2.5 Pro