Subaxis started as an API gateway with one simple goal: make LLM access affordable enough that anyone could use it — students, indie devs, small teams, not just companies with big budgets.
And for a while, it worked. Better than we expected, honestly. At our peak we offered a catalogue of 300+ models and had more than 2,000 active users building on top of us every day. Behind the scenes we built and ran our own inference engine, which let us deliver high-quality output, fast response speeds, and lower inference costs — savings we passed straight on to our users.
But growth is a double-edged sword. Serving this many users takes serious compute, and right now we simply can't keep up with the demand. Rather than quietly degrade the service or start cutting corners on quality, we're choosing to shut things down properly.
Your balance doesn't disappear with us. You can still access it through a private panel — just open a ticket on our Discord server and request access, and we'll sort you out:
discord.gg/CyACVYFWsT
To everyone who used Subaxis, reported a bug, or just trusted us with their API calls — thank you. It meant a lot.