@cf/google/gemma-7b-it-lora

Streams a reply to a ~20-word prompt, capped at 100 tokens; latency is the full reply.