ByteBrief
We're a portrait publication through and through. Turn your phone back and your briefing picks up right where you left it.
(We tried widescreen once. It wasn't us.)

Xiaomi MiMo lead Fuli Luo said MiMo-V3 will adopt a new architecture built on HySparse2, which cuts long-context inference costs and improves retrieval for agentic workloads. At 1 million tokens, HySparse2 reduces prefill computation 5.02 times and KV cache 4.5 times, with higher MRCRv2 and RULER-v2 scores and lower AgentPPL and LongPPL results.
Tap to vote and see what everyone thinks.
Summary by ByteBrief