Xiaomi's MiMo-V2.6-Pro debuts as the top-ranked open-weights AI model, overtaking DeepSeek

Xiaomi's MiMo-V2.6-Pro debuts as the top-ranked open-weights AI model, overtaking DeepSeek

Xiaomi released MiMo-V2.6-Pro, which the company says is now the highest-performing open-weights model available worldwide, ahead of DeepSeek. Alongside the flagship, Xiaomi shipped MiMo-V2.6-Flash, a cheaper variant aimed at lower-cost inference. Both ship as open weights, extending Xiaomi's presence in a foundation-model race that until now has been dominated by names like DeepSeek and Alibaba's Qwen line.

The release matters because it puts a hardware-first company squarely into the open-weights leaderboard conversation, the same conversation DeepSeek reshaped when it forced a rethink of what open, low-cost training can achieve. A cheaper Flash tier alongside a claimed-best Pro tier suggests Xiaomi is trying to compete on both capability and price simultaneously, a two-tier playbook other Chinese labs have used to widen adoption. Per the AI Market Watch index, which tracks Xiaomi under its Robotics/Embodied AI coverage following its prior open embodied-AI model releases, our pipeline actually logged fewer Xiaomi-tagged stories in the last 90 days (5) than in the 90 days before (8) — a reminder that a single benchmark claim doesn't necessarily mean sustained attention, and that this count reflects ingested sources rather than a full census.

For builders, a credible open-weights alternative from a company with Xiaomi's device distribution lowers the cost of experimenting with a top-tier model and adds pricing pressure at the low end via the Flash tier. For investors, the open question is whether Xiaomi treats MiMo as a commercial API business or as an ecosystem asset tied to its hardware and embodied-AI roadmap; the answer determines whether this is a new revenue line or a strategic subsidy.

#Xiaomi #MiMoV2.6 #OpenWeights #DeepSeek #FoundationModels #ChineseAI