Hey Product Hunt! 👋 I'm building 99 because I was tired of watching my AI API bills spike every time I ran more agents or longer eval loops. The problem is simple: the more useful your AI agents become, the more expensive they get. I found myself shrinking prompts, skipping evals, and checking usage dashboards like a bank account. So I'm building 99 — a flat-rate AI API powered by NVIDIA Nemotron 3 Super 120B (MoE, 12B active). One price, unlimited requests, OpenAI-compatible. What you'll get: — 120B parameter model, 128K context per request — 16 concurrent requests — OpenAI-compatible (just swap your base URL) — Early bird: $69.9/mo locked forever The POC is done. I've already benchmarked real workloads — 16 agents running simultaneously, 128 requests at 100% success rate, multi-turn sessions, burst traffic patterns. The infra works. Now I need to know if enough builders actually want this. No charge until we go live. Your signup is what decides if this ships. Reserve a seat — it takes 30 seconds. What matters most to you in an AI API: speed, cost, or model quality?

99 — Unlimited AI API

One flat price. Unlimited AI API. Never nervous.

One flat price. Unlimited AI API. Never nervous.