architectures · openbmb

MiniCPM SALA

Dense decoder-only transformer built for very long context: 512K positions with an aggressive 16:1 grouped-query ratio (32 query heads against 2 key/value heads), which is what keeps the KV cache tractable at that length — cache size scales with key/value heads, not query heads. Declared as model_type `minicpm_sala`.

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

Publisher	openbmb
Origin	China
Licence	Apache-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.