Xiaomi MiMo-V2.6-Pro Leads Open-Model Rankings, Anthropic Alleges Distillation
Xiaomi released MiMo-V2.6-Pro with 1 trillion parameters and 42 billion active per request; it scored 72.6 on DeepSWE. Anthropic accuses the company of distilling Claude through more than 400,000 queries.
Xiaomi released MiMo-V2.6-Pro, an open model with 1 trillion parameters, 42 billion of which are active per request. On the DeepSWE benchmark it scored 72.6, up from the previous 58.4. It now leads the open-model rankings on the Intelligence Index with 46 points. Training cost $2.62 million and took fewer than six days.
In a threat report, Anthropic described case GTG-16008. Over 20 days in March and April 2026, Xiaomi ran more than 400,000 conversations through Claude, reportedly to distill the responses into MiMo. The question of where that quality comes from is the first one I would ask before putting this model into a production pipeline.
Source: the-decoder.com
Free course
Stop reading about AI — start building with it
The free Claude Code course: your first site, tool or game — no coding. No upsells, no cross-sells — nothing to buy here.
Start free →
Author
Evgenii Arsentev
PhD · Chief Executive Officer, digital health
Articles · Latest articles