Baidu open-sources ERNIE 5.5 Flash Base: 280B params, 14B active
BaiduResearch · x · 2026-11-05
Baidu Research open-sourced ERNIE 5.5 Flash Base, a 280B-parameter model with 14B active per token, claiming SOTA at its size, with the full ERNIE 5.5 Flash on the way. Key efficiency designs: Latent-Gated Sparse Attention concentrating KV-cache on few full-token layers, Multi-Router Latent MoE with router voting, ECHO multi-token prediction doubling as a free speculative-decoding draft head, plus a distributed Muon optimizer and FlashMask-4 attention kernel.
More from Infra
- Apple's coreai-models repo (2.2k stars) now accepts third-party contributions — pcuenq · 2026-11-05
- oMLX lets a second Mac boost prefill up to 2.9x on big-memory Macs — pcuenq · 2026-11-05
- FineWeb-30M speedrun record falls to 14.2s on one 8xH100 node, a 1.64x speedup — RobertTLange · 2026-11-05
- HydraDB: a Rust graph database that keeps state in S3 and treats query workers as disposable compute — JeremyCMorgan · 2026-11-05
- OpenSSH 10.6p2 released with a workaround for a critical vulnerability — Animux · 2026-11-05
- Microsoft rejects multibillion-dollar AI deal with UK's Nscale, FT reports — nordicinst · 2026-11-05