architectures · sarvamai

Sarvam MLA

Multi-head latent attention combined with sparse experts: keys and values are projected into a shared 512-dimension latent (kv_lora_rank 512) rather than cached per head, alongside 128 experts with 8 routed per token. Because MLA compresses KV into a shared latent, the standard per-head KV-cache formula does not apply to this architecture and the registry declines to estimate rather than reporting a wrong number. Declared as model_type `sarvam_mla`.

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	sarvamai
Origin	India
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.