architectures · arcee-ai

AFMoE (Arcee Foundation MoE)

Sparse mixture-of-experts transformer with a wide expert pool and narrow routing — 256 experts with 4 routed per token — combined with grouped-query attention and a 4096-token sliding window. The wide-pool/narrow-route ratio is the design's distinguishing choice: capacity scales with the pool while per-token compute stays near a much smaller dense model. Declared as model_type `afmoe`.

Verify

What this registry has checked

Quarantine. We hold the publisher’s own published hash for these weights and have recorded it against this entry. We have not downloaded and hashed the weights ourselves, so the hash is the publisher’s claim, not our measurement.
Re-check this yourself → Source → Full record →

What is recorded

Publisherarcee-ai
OriginUnited States
LicenceApache-2.0

Elsewhere in the registry

All architectures → Browse with filters → What will run on my hardware? →

Registry record as of 2026-08-14. Identity, licence and parameter count are properties of a release and do not move; tier can, and the verify link above re-checks it against the live registry rather than this page.