Blazing-fast open AI models — your own private key, a few dollars a month.
What you get
Serious speed. A 120-billion-parameter model on hardware that streams answers faster than you can read them.
Works in the tools you already use. Cursor, aider, or any app/SDK that speaks the OpenAI API — two settings and you're in.
Your own key, your own budget. Personal API key with a fixed monthly dollar allowance. No surprises, no shared accounts.
Private by design. Prompts and responses are never stored — the service only counts tokens to meter your budget.
A model for each job. A fast flagship, a premium coder, a vision model that reads images, and embeddings — with more added as we grow.
Pricing
Tier
Monthly allowance
Good for
$2
light use
casual chat, occasional questions
$5
everyday use
daily chat + coding — most people never hit it
$10
heavy use
agents, long coding sessions, premium models
Allowances are metered per token at real cost — you pay for what the models
actually burn, nothing else.
Invite-only.
Ava AI runs on shared infrastructure for a small circle of friends and colleagues.
Want in? Message Venkata SR — vs dot rajanala plus ava at gmail dot com — setup takes two minutes.
Fair use: each key has rate limits (30 requests/min, 2 parallel) so one
person can't slow everyone down. Keep your key private — anyone holding it
spends your budget.