MiMo v2 Flash vs MiMo v2.6 Flash
Different efforts: non-thinking · thinking
Humanity's Last Exam Humanity's Last Exam: Very hard expert questions across many fields, written to stump today’s models
MiMo v2 Flash8.6
MiMo v2.6 Flash35.1
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
MiMo v2 Flash—
MiMo v2.6 Flash22.7
AutomationBench-AA AutomationBench-AA: Agents orchestrate SaaS workflows through APIs; objectives receive credit only when the task has no guardrail violation
MiMo v2 Flash—
MiMo v2.6 Flash64.1
Arena Text · creative writing Arena Text · creative writing: Human pairwise preferences in the stated text vote pool
MiMo v2 Flash1,353
MiMo v2.6 Flash1,390
MMMU-Pro (vision) MMMU-Pro (vision): College-level questions that require understanding images: charts, diagrams, photos
MiMo v2 Flash—
MiMo v2.6 Flash73.1
| Benchmark | MiMo v2 Flash | MiMo v2.6 Flash |
|---|---|---|
| Humanity's Last Exam | 8.6 | 35.1 |
| Terminal-Bench 4.0 | — | 22.7 |
| AutomationBench-AA | — | 64.1 |
| Arena Text · creative writing | 1,353 | 1,390 |
| MMMU-Pro (vision) | — | 73.1 |
Specs
| Spec | MiMo v2 Flash | MiMo v2.6 Flash |
|---|---|---|
| Input, per 1M | $0.14 | $0.14 |
| Output, per 1M | $0.28 | $0.28 |
| Cost per task? | — | $0.062 |
| Speed, tok/s | — | 41 |
| Context | 256K | 1M |
| Weights | Open | Open |
| Released | Dec 2025 | Sep 2026 |
| Memory (4-bit) | 186 GB | 186 GB |
| Licence | MIT | MIT |
| Commercial use | Allowed | Allowed |
| See images | — | ✓ |
| Read PDFs | — | — |
| Hear audio | — | ✓ |
| Watch video | — | ✓ |
| Call tools | ✓ | ✓ |
| Force a tool call | ? | Partial |
| Return strict JSON | Partial | ✓ |
| Search the web | ? | ? |
| Run code | ? | ? |
| Use a computer | ? | ? |
What breaks when you switch
MiMo v2 Flash → MiMo v2.6 Flash
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 Flash → MiMo v2 Flash