MiMo-V2.6-Flash dropped yesterday.
This is a AI post classified by Jev as Frontier models (a model release), kept by the AI Radar because it carries real work, not commentary.
MiMo-V2.6-Flash dropped yesterday. I ran my full grid on it, on my two DGX Sparks. Here is what it is and how it runs 👇 WHAT IT IS Xiaomi's efficiency tier of the V2.6 series. 309B sparse MoE, 15B active per token, ships natively in mixed precision: MXFP4 experts, FP8 attention. 1M native context. Text, image, video and audio input in one checkpoint, with a bundled 5-layer DFlash-style drafter doing 7-token speculative decoding out of the box. One mixed RL run across coding, agents, vision and security. They call it "You Only RL Once." A self-improvement loop where an agentic grader ran
Posted by Yume_X (2.1k followers) 59 min ago · 0 likes · 43 views · view the original post on X. Kept by the AI Radar as Frontier models.
More AI work like this
- GPT-6 Sol, Luna, Opus 5.5, and Grok 4.7 are in the Playground — @skalskip92
- Testing animations with Claude Opus 5.5 on a recent essay of mine. This sample came from… — @tracewoodgrains
- We created this game-style pixel-art animation in After Effects with Claude Opus 5.5 +… — @higgsfield_ai
- Claude Opus 5.5 vs GPT-6 Sol: both were asked to make prosthetics more accessible by… — @higgsfield_ai
- Claude Opus 5.5 & GPT6 Sol 性能調査！ — @gimu_ai
- Opus 5.5 — @Perpetualmaniac
- DrivingBench benchmarks frontier LLMs by giving them direct control of a Toyota Corolla — @0x0SojalSec
- Great to see Anthropic reporting biomedical image analysis capabilities... — @iScienceLuvr
Every post is read and classified by Jev (TypeSafe): what it is, which market it belongs to, and whether the link is a real tool. 43.6k posts from 5k X accounts over the last 14 days, 1.8k tools, 19 markets. Collected every 5 minutes, fully re-ranked every hour — last update 2026-09-23 00:23 UTC. Full method.