ERNIE 5.5 Flash Preview: 280B-param multi-router MoE activating only 14B per token
xiaohu · x · 2026-11-11
ERNIE 5.5 Flash Preview is a multi-router MoE: 280B total parameters with only 14B activated per token, and multiple routers jointly select experts for better routing at the same activation budget. It uses hybrid sparse attention — most layers read compressed history, a few layers retrieve details — saving memory on long contexts.
A companion demo shows two agents building an airport and planes in Blender via the Python API: one researches maps and builds terminals/runways, the other models fuselage and wings with stock cockpit/cabin assets; they then reconcile scale and coordinates, and render with Cycles/OptiX before compositing via FFmpeg.
More from Models
- Google open-sources Harness², an agent that rewrites its own setup per task — Arindam_1729 · 2026-11-11
- Dev dumps Codex desktop app over quality issues, falls back to codex-cli — tekbog · 2026-11-11
- Why do old models get dumber before launches? Lab party, says joking X post — glcst · 2026-11-11
- Grok Bot now detects hidden prompt injections, warns users and refuses to follow them — XFreeze · 2026-11-11
- Shanghai Institute Open-Sources Nex-N2.6, a 321B Multimodal Agentic Model — victormustar · 2026-11-11
- Nex-N2.6 ships open FP8 weights (321B, GLM-5.3-Flash based), cuts reasoning loops 16%→1.6% — Physical_Hat4022 · 2026-11-11