Mistral: Mistral Large 4 vs Xiaomi: MiMo-V2.6-Pro-UltraSpeed
Side-by-side comparison of specs, pricing, benchmark scores, and task rankings. Updated 2026-10-07.
| Mistral: Mistral Large 4 | Xiaomi: MiMo-V2.6-Pro-UltraSpeed | |
|---|---|---|
| Vendor | mistralai | xiaomi |
| Quality Score | 100 | 100 |
| Input Price | $0.68/M | $4.35/M |
| Output Price | $2.09/M | $8.70/M |
| Context Window | 524,288 | 1,048,576 |
| Max Output | 262,144 | 131,072 |
| Tool Calling | ✓ | ✓ |
| Structured Output | ✓ | ✓ |
| Reasoning Mode | ✓ | ✓ |
| Vision | ✓ | ✓ |
| Audio | - | ✓ |
The verdict
For a sample job of 1 million input tokens and 250,000 output tokens, Mistral Large 4 costs $1.20 and MiMo-V2.6-Pro-UltraSpeed costs $6.52, so Mistral Large 4 is about 5.4 times cheaper for the same work.
MiMo-V2.6-Pro-UltraSpeed reads up to 1,048,576 tokens at once, roughly 1,573 pages of text, against about 786 pages for Mistral Large 4.
Only MiMo-V2.6-Pro-UltraSpeed offers audio input and video input.
Across the 10 tasks where they differ, Mistral Large 4 scores higher on 10 (SQL Generation, Code Completion, Code Documentation and others).
Mistral Large 4 was released in October 2026 and MiMo-V2.6-Pro-UltraSpeed in September 2026.
Pick Mistral Large 4 if you need lower cost or the stronger all-round task scores; pick MiMo-V2.6-Pro-UltraSpeed if you need long documents, audio input or video input.
Costs use current list prices per million tokens; page counts assume about 0.75 words per token and 500 words per page.
Who wins by task?
| Task | Mistral: Mistral Large 4 | Xiaomi: MiMo-V2.6-Pro-UltraSpeed |
|---|---|---|
| SQL Generation | 133 | 132 |
| Code Completion | 131 | 117 |
| Code Documentation | 131 | 129 |
| Regex Writing | 119 | 117 |
| CSV / Spreadsheet Cleanup | 133 | 132 |
| JSON Extraction | 131 | 121 |
| Bulk Data Labeling | 129 | 118 |
| Long-Document Summarization | 137 | 136 |
| Short-Form Summarization | 123 | 113 |
| Blog Post Writing | 121 | 120 |
Scores reflect capability match + benchmark data + pricing for each task. Methodology →