Xiaomi joined China's AI price war this week, cutting prices across its MiMo-V2.5 model family by as much as 99%. The reductions are permanent and took effect May 27, per the company's pricing announcement. Xiaomi also dropped the tiered pricing that scaled with context length.
For the flagship MiMo-V2.5-Pro, third-party listings on OpenRouter show $0.435 per million input tokens and $0.87 per million output. That 99% headline leans on prompt caching, though. Cache-hit input falls to a fraction of a cent, while the cut on standard uncached input lands closer to 78%.
The timing isn't subtle. DeepSeek pulled a similar move last week, making permanent a cut that drops its V4-Pro to a quarter of the old price. Both are chasing enterprise developers who burn through billions of tokens on agentic workloads.
Xiaomi reworked its subscription Token Plan too, lifting usage limits five to eight times and resetting every active subscriber's credit balance. "Enabling more people to use better models," per the announcement, which is the expected mission-statement line. The Pro model, an open release, claims parity with Western frontier systems on benchmarks like SWE-bench Pro, though those comparisons are Xiaomi's own.
Bottom Line
MiMo-V2.5-Pro now lists at $0.435 per million input tokens and $0.87 per million output on OpenRouter.
Quick Facts
- MiMo-V2.5-Pro: $0.435 per million input tokens (OpenRouter listing)
- MiMo-V2.5-Pro: $0.87 per million output tokens
- Price cuts up to 99%, effective May 27, 2026 (Beijing time)
- Context window: 1,048,576 tokens
- Token Plan usage limits raised 5 to 8 times (company-reported)
