A Classification Model That Runs on a Phone: Supersonic Labs Ships Julia-1 and Reopens the 'Do You Even Need an LLM' Debate
Brazil's Supersonic Labs released Julia-1, an Apache 2.0 classification model built on mmBERT that runs on CPUs, tablets, and smartphones — and priced token access at a fraction of a cent, reviving a growing argument that many production tasks never needed a large language model at all.
Supersonic Labs, a Brazilian AI lab, introduced Julia-1 this weekend with a deliberately modest pitch: "Our first classification model that runs on almost anything," the company wrote. In a discourse dominated by trillion-parameter frontier systems, that sentence is the news. Julia-1 is not trying to reason, write code, or hold a conversation. It classifies and ranks — and it does so, per the lab, without a GPU in sight.
The technical shape matters. According to @AGTPinsights, Julia-1 is built on mmBERT, released under an Apache 2.0 license, and runs across CPUs, tablets, and phones. Pricing was pegged at roughly $0.025 per million input tokens — an order of magnitude cheaper than what teams have grown accustomed to paying frontier labs for equivalent classification work routed through a general-purpose model. The permissive license removes the usual friction: teams can inspect it, fine-tune it, and deploy it on their own hardware without a per-seat contract.
Get our free daily newsletter
Get this article free — plus the lead story every day — delivered to your inbox.
Want every article and the full archive? Upgrade anytime.
No spam. Unsubscribe anytime.