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