Most AI APIs charge per token, which means your bill scales with usage. That's fine for prototyping, but at production scale it becomes unpredictable. A single bad prompt loop can spike your bill 10x overnight.
Plugsky's self-serve plans include every model in your tier for a flat monthly price. You commit to the price; published daily limits keep capacity predictable.
Why predictable limits matter
- Predictability: you know the bill before you ship. No meter, no surprises.
- No "free tier that throttles at peak." Real production traffic, not toy demos.
- Encourages innovation: when every call is metered, teams avoid experimentation. When per-token meters are absent, teams iterate faster.
- Better unit economics at scale: once you're past ~20M tokens/month, flat pricing is cheaper than per-token on every public AI vendor.
What's included in each tier
Every plan has published fair-use limits — no per-token meter, no surprise bills. RPM, concurrency and max input/output caps protect shared capacity.
| Plan | Monthly | Models in tier | Daily requests / tokens | Rate limit (req/min) |
|---|---|---|---|---|
| Hobby | $8 | micro → pro | 1,000 req/day · 2M tokens | 120 |
| Starter | $20 | micro → pro | 2,000 req/day · 5M tokens | 240 |
| Builder | $60 | + max, vision | 5,000 req/day · 15M tokens | 600 |
| Scale | $120 | + frontier, reasoning | 10,000 req/day · 30M tokens | 1,200 |
What are the limits? Daily requests and tokens, concurrency, RPM and max input/output are published on the pricing page. At 80% the API returns an X-Plugsky-Quota-Warning header; at 100% it returns a clear 429 with an upgrade link. Embeddings, RAG queries and agent runs share the same plan allowance — no per-feature add-ons.
When per-token is better
Per-token pricing beats flat pricing only at very low usage. If you're under ~5M tokens/month, you're probably better off on plugsky-micro (free) or the $20 Starter plan with the same effective economics.
Past 50M tokens/month, flat pricing on Plugsky Builder ($60) saves you 60-90% vs. OpenAI gpt-4o for the same workload. Use the calculator for your exact numbers.
Frequently asked questions
Is there really no per-token meter?
Correct, there is no per-token meter. Self-serve plans have published daily request and token caps and 120/240/600/1,200 RPM.
What about embeddings and RAG?
Same — covered by the plan's daily limits and fair-use policy. Embeddings, RAG queries, agent runs and tool calls share the same flat pricing.
What if I exceed the rate limit?
You get a 429 with a Retry-After header. Upgrade to the next tier for higher rate limits, or contact sales for custom limits.
Is there a fair-use policy?
Yes — self-serve plans have published daily request/token caps plus RPM, concurrency and max input/output limits, and a fair-use policy against abuse or scraping. Enterprise contracts have custom limits.
Start free
Start free — then Starter is $20/mo flat with published fair-use limits, 38 models.
Start Free → See full pricing