Xiaomi Open-Sources MiMo-V2.6-Distill-Qwen-9B with 7,000+ RL Environments
Xiaomi's MiMo team released MiMo-V2.6-Distill-Qwen-9B on Hugging Face on September 22, distilling MiMo-V2.6 onto the Qwen3.5-9B architecture along with a tech report and 7,000+ RL task environments. The model climbed Hugging Face trending charts. It is widely seen as another instance of official distillation from a large model to a small architecture, echoing DeepSeek R1's distillation playbook.
Confirmed
- The model is a 9B-parameter agentic model built on Qwen3.5-9B, trained via distillation and SFT with MiMo self-generated data, offered in Pro and Flash versions.
- It targets agentic, coding, and tool-use capabilities as an image-text-to-text model.
- The official @XiaomiMiMo account confirmed the model reached Hugging Face trending.
- Per @tekbog, the release also includes a tech report, 7,000+ RL task environments, an end-to-end RL framework, and composable mini-harnesses for the community.
Unconfirmed
- No third-party benchmarks were included in the posts; actual performance awaits the model card and community testing.
- @teortaxesTex notes the model has not undergone any independent RL yet, so post-RL performance remains to be seen.
Why it matters
- @victormustar argues the official distill variant recreates DeepSeek R1's classic Qwen/Llama distillation route, making it more promising than community distills for users wanting small reasoning models.
- @teortaxesTex points out the 9B base is already strong and improves further with MiMo SFT data, while the open RL environments give the community infrastructure to run its own RL.
- @tekbog notes users jokingly called out OpenAI in response, reflecting positive community sentiment toward Xiaomi's release scale.
2026-09-22 ~ 2026-09-24 · 8 related posts
Primary sources
- [source] Xiaomi Releases Small Qwen 3.5 9B Distill SFT'd on MiMo Data, Plus RL Environments — teortaxesTex · 2026-09-22
- Xiaomi distills MiMo-V2.6 into an open-source Qwen-based 9B model on Hugging Face — Beamsters · 2026-09-22
- [source] Xiaomi's MiMo-V2.6-Distill-Qwen-9B trends on Hugging Face — XiaomiMiMo · 2026-09-22
- Xiaomi Releases MiMo-V2.6-Distill-Qwen-9B, a 9B Agentic Model Distilled from Qwen3.5 — anovers · 2026-09-22
- Xiaomi MiMo V2.6 Ships Official Distill-Qwen-9B Variant, Echoing DeepSeek R1 Distills — victormustar · 2026-09-22
- Xiaomi Releases MiMo-V2.6-Distill-Qwen-9B Distilled Model — Aggravating-Push-207 · 2026-09-22
- [source] Xiaomi open-sources MiMo-V2.6 distill models and 7K+ RL environments, drawing OpenAI comparisons — tekbog · 2026-09-24
1 near-duplicate retellings: victormustar