After DeepSeek V4.1 Flash, Xiaomi MiMo is moving to a causal encoder-decoder for its…
This is a AI post classified by AGK Intelligence as Frontier models (a model release), kept by the AI Radar because it carries real work, not commentary.
After DeepSeek V4.1 Flash, Xiaomi MiMo is moving to a causal encoder-decoder for its next-gen models. HySparse2: prompts run through only 25 of 49 layers. Decoder KV is projected from encoder states. At 1M context vs MiMo-V2: 5x less prefill compute, 4.5x smaller KV cache.
Posted by SemiAnalysis (183.1k followers) 1 h ago · 125 likes · 9.6k views · view the original post on X. Kept by the AI Radar as Frontier models.
More AI work like this
- FrontierCS 2 update: 4 new models join the leaderboard. Claude Sonnet 5.5 beats the… — @MangQiuyang
- 😂DeepSWE高分集体翻车！MaxxBench一测水分全被取挤掉了？！ — @NFT_Chen
- Terra has the flattest speed variance curve of any OpenAI model — @AdamHoltererer
- Since i don't have access to Qwen4, decided to burn through my DeepSeek credits !!! — @ItsmeAjayKV
- Qwen Qoduck（Qwen 4?） 超惊艳 3D 白宫建模测试作品欣赏： — @leo114119
- Devin CLI 3000.12.17有泄漏了Claude的模型。。。。 — @mylifcc
- The only model that actually scored higher on DeepSWE when small changes were made to… — @deedydas
- New OpenAI and Anthropic model entries have surfaced inside Devin CLI. — @Smlay_ero
Every post is read and classified by AGK Intelligence: what it is, which market it belongs to, and whether the link is a real tool. 65.2k posts from 5k X accounts over the last 14 days, 10.6k tools, 19 markets. New posts are collected several times an hour and the full ranking is recomputed every hour — last update 2026-11-07 20:50 UTC. Full method.