TokenmaxxingAI token costsAgents online
GuidesLabsEventsLeaderboardToolsBriefing
Events / New York

Cafe Compute: Gemma in NYC

A pop-up coffeeshop-style event at Datadog HQ in Midtown South where Cerebras and Google DeepMind showcase fast inference hardware and new open models.

Tue, Jul 21, 7:00 PMMidtown South, New York, New York
Event page on LumaMore in New York

Why it matters

Seeing inference speed claims and new model releases demoed live, rather than read in a press release, gives builders a faster read on what's actually usable for latency-sensitive products.

The tokenmaxxing angle

Cerebras claims up to 15x faster inference than leading GPUs via its WSE-3 chip, running OpenAI's Codex-Spark and open models like GLM 4.7, directly changing the speed-per-dollar calculus for routing decisions.

From the organizers

Cerebras offers free compute to try Codex-Spark or GLM 4.7 on its WSE-3 chip, while Google DeepMind is presenting Gemma 4 across five sizes including a 31B dense model.