Xiaomi MiMo-V2.6 Flash vs Qwen 2.5 Max
A direct, empirical evaluation of Xiaomi MiMo-V2.6 Flash (Xiaomi AI Lab) against Qwen 2.5 Max (Alibaba Cloud).
Xiaomi MiMo-V2.6 Flash
Developed by Xiaomi AI Lab · 185 Billion (Sparse MoE) parameters
Low-latency agent loops, high-volume classification, and developer code completion.
Qwen 2.5 Max
Developed by Alibaba Cloud · Proprietary (~700B) parameters
Cross-lingual document extraction and structured structured output generation.
| Metric | Xiaomi MiMo-V2.6 Flash | Qwen 2.5 Max |
|---|---|---|
| SWE-bench Verified | 64.2% | 51.4% |
| MMLU-Pro | 82.4% | 77.2% |
| MATH-500 | 89.5% | 89.4% |
| Context Window | 128,000 tokens | 128,000 tokens |
| Input Pricing (per 1M) | $0.04 | $1.60 |
| Output Pricing (per 1M) | $0.08 | $6.40 |