--- format: "aidr-story-markdown/v1" id: "f1dca09efa5746ae7cfd7149212ef989f7919d43df80469d0d831fa076809d3e" canonical_url: "https://aidr.today/f1dca09e?lang=en" title: "Xiaomi Cuts MiMo-V3 Prefill Compute 5.02x With HySParse2 Architecture" lang: "en" requested_lang: "en" available_langs: ["en","vi"] translation_fallback: null fallback_fields: [] published_at: "2026-09-23T17:17:53.000Z" category: "Research" topics: ["xiaomi","inference","agent","chips","infra"] source_urls: ["https://marketbrief.now/ai/update-xiaomi-cuts-mimo-v3-prefill-compute-502x-with-hysparse2-architect-1fc708bb","https://x.com/_LuoFuli/status/2102766365190901957","https://x.com/teortaxesTex/status/2102775394206179335","https://huggingnews.com/ai/update-xiaomi-cuts-mimo-v3-prefill-compute-502x-with-hysparse2-architect-1fc708bb"] summary: "The computational demands of agentic inference are at the center of a new design released by Xiaomi for its next-generation AI models. The HySparse2 architecture, which will serve as the core of MiMo-V3, achieves 5.02x lower prefill FLOPs and a 4.5x smaller KV cache at 1M tokens compared to the Hybrid SWA architecture used in MiMo-V2.6. This update also improves long-context retrieval as measured by MRCRv2 and RULER-v2 scores while lowering AgentPPL and LongPPL. The design shifts to optimize for workloads where short actions return long observations that require frequent prefilling of a growing context. To reduce overhead, HySParse2 utilizes 'KV Bridging' to build cross-decoder K/V from self-decoder hidden states and 'KV Reuse' for sparse layers. Additional changes include moving to token-level selection and implementing a forced window of recent tokens to allow local and global tokens to share a single KV cache." --- # Xiaomi Cuts MiMo\-V3 Prefill Compute 5\.02x With HySParse2 Architecture > [Open the canonical story]() **Published:** 2026-09-23T17:17:53.000Z **Category:** Research **Topics:** xiaomi, inference, agent, chips, infra ## Summary The computational demands of agentic inference are at the center of a new design released by Xiaomi for its next\-generation AI models\. The HySparse2 architecture, which will serve as the core of MiMo\-V3, achieves 5\.02x lower prefill FLOPs and a 4\.5x smaller KV cache at 1M tokens compared to the Hybrid SWA architecture used in MiMo\-V2\.6\. This update also improves long\-context retrieval as measured by MRCRv2 and RULER\-v2 scores while lowering AgentPPL and LongPPL\. The design shifts to optimize for workloads where short actions return long observations that require frequent prefilling of a growing context\. To reduce overhead, HySParse2 utilizes 'KV Bridging' to build cross\-decoder K/V from self\-decoder hidden states and 'KV Reuse' for sparse layers\. Additional changes include moving to token\-level selection and implementing a forced window of recent tokens to allow local and global tokens to share a single KV cache\. ## Sources - [Story source]() - [Story source]() - [Supporting source]() - [Story source]()