Xiaomi (小米) permanently cuts MiMo‑V2.5 API prices, resets Token Plan quotas

What happened

Xiaomi (小米) said it has permanently overhauled pricing for its MiMo‑V2.5 series API, with it has been reported that new prices represent reductions ranging from roughly 57% up to 99% versus the original list. The company also removed pricing distinctions based on context window length. The changes took effect globally at 00:00 Beijing time on May 27, according to the announcement.

Who is affected and how

Token Plan billing has been reworked as well: Xiaomi says effective usage per credit will rise roughly 5–8× under the new rules, and the charging logic has been simplified. It has been reported that all currently active Token Plan subscribers — including users who received Token Plans through the "百万亿 Token 创造者激励计划" (the “Hundreds‑of‑Trillions Token Creator Incentive Program”) and Apache Software Foundation beneficiaries — will see their Credits fully reset at 00:00 Beijing time on May 27 and be charged under the new scheme. Xiaomi also said past, expired Token Plan payers will receive a surprise offer to be announced within a week. It has been reported that the 100T tokens allocated for the creator incentive were fully distributed ahead of schedule by May 26, 16:08 Beijing time.

Why it matters

For Western readers: MiMo is Xiaomi’s multimodal AI model family and these API moves are meant to drive broader developer adoption by slashing operating costs. Why now? Analysts and observers point out the cut comes as Chinese cloud and AI vendors race to scale services domestically amid tighter Western export controls on advanced chips and model components — price and accessibility are natural levers. Will this spark a price war among Asia’s AI providers or simply accelerate adoption by startups and enterprise customers? Either way, lower API bills make running large‑model workloads materially cheaper for customers inside China and internationally.