Why it matters
Open-weight models like Gemma let builders control inference costs by self-hosting. A hackathon produces working prototypes and surfaces real cost/performance tradeoffs between managed and self-hosted endpoints.
The tokenmaxxing angle
Gemma is self-hostable: the cost baseline is compute, not per-token API pricing. Teams discover where Gemma throughput and context limits push them toward or away from hosted API alternatives.
From the organizers
Host: Ramin Madi, PhD, Northeastern adjunct faculty. Venue: Curry Student Center, 360 Huntington Ave, Boston. Partner: Northeastern University. June 26, 2026, 9AM start; prizes mentioned.