26M parameter function call model that runs on incredibly small devices with 6000 toks/sec prefill speed
Low-latency AI engine for mobile devices and wearables with hybrid edge-cloud computing
