Key facts
| Free plan | 2 free AI models (plugsky-micro and plugsky-lite), 2 API keys, no card |
| Self-serve plans | Flat monthly with unlimited fair-use usage; no per-token charges or overage fees |
| Trial | 14-day full-access trial for evaluating production-like workloads |
| Model access | 30+ models from small and fast to frontier reasoning under one API |
| Enterprise | Committed capacity, negotiated limits, SLA, SSO, VPC, on-prem and air-gapped |
| Rate limits | RPM, TPM and concurrency caps apply instead of token meters |
| Compatibility | OpenAI-compatible API — change the base URL and model name |
| Product status | Chat, streaming, JSON mode, function calling, embeddings and RAG are live |
TL;DR
- Free plan: two models, two API keys, no card required.
- Paid self-serve: flat monthly, unlimited fair use, no per-token billing.
- 14-day full-access trial for real workloads before you commit.
- One API and one plan cover 30+ models.
- Enterprise adds committed capacity, SLA and sovereign deployment options.
How it works, step by step
- Start on the free plan and confirm the API works with your existing SDK.
- Measure tokens and requests per task during development.
- Use the 14-day full-access trial to test production-like workloads.
- Compare measured usage with self-serve plan limits and headroom.
- Choose the self-serve plan that covers peak usage with at least 1.5x headroom.
- Talk to the team about enterprise terms if you need guaranteed capacity or residency.
- Review usage analytics monthly and adjust the plan as traffic grows.
Original data
Try it yourself
Open the AI model pricing calculator →
The free plan: two models, no card
The free plan is designed for real prototyping, not a trial that expires. It includes two AI models — plugsky-micro and plugsky-lite — plus two API keys, and it requires no credit card. Both models are reachable through the same OpenAI-compatible endpoint as paid plans, so code written on the free plan keeps working when you upgrade.
Use it to validate prompts, latency and integration effort. When you need frontier models, higher limits or production headroom, the paid plans and the 14-day full-access trial are the next step.
Self-serve plans: flat-rate, fair use
Paid self-serve plans are flat monthly. There are no token meters, no per-token charges and no overage fees — usage is unlimited under fair use, and RPM, TPM and concurrency limits protect shared capacity. That means the bill does not move when a prompt grows, a retry fires or an agent loop takes an extra step.
Prices change over time, so this article never quotes them: see the live pricing page for current plans and limits. The pricing calculator helps compare your measured usage with plan tiers.
One plan, 30+ models
Model access is not priced per model on self-serve plans. The catalogue spans 30+ models from small, fast options for classification and formatting to frontier models for reasoning, plus embedding models for retrieval. Routing between them is a code decision, not a procurement one.
That structure lets you optimise for quality and latency: send easy requests to small models, reserve frontier capacity for hard ones, and change routing without renegotiating anything.
Enterprise: capacity, residency and SLA
Enterprise agreements are for teams that need guarantees rather than best effort. They add committed capacity with negotiated limits, a published SLA, SSO and RBAC, audit logs, DPA and BYOK options, and deployment choices including VPC, on-prem and air-gapped environments for regulated regions.
The trigger to move is simple: when you start planning launches around rate limits, or when procurement needs contractual residency and uptime commitments, enterprise terms are the right home. Until then, self-serve flat-rate covers normal production traffic.
Honest comparison
| Plan aspect | Free | Self-serve paid | Enterprise |
|---|---|---|---|
| Cost model | Free, no card | Flat monthly, unlimited fair use | Committed capacity |
| Models | plugsky-micro and plugsky-lite | 30+ models under one plan | 30+ models, negotiated catalogue |
| Token billing | None | None — no per-token charges or overage fees | Committed terms |
| Limits | Fair-use free-tier limits | Higher fair-use RPM, TPM, concurrency | Negotiated and guaranteed |
| Deployment | Plugsky cloud | Plugsky cloud | VPC, on-prem, air-gapped |
| Best for | Prototyping and evaluation | Production applications | Regulated and heavy workloads |
Frequently asked questions
What does the Plugsky free plan include?
Two free AI models, plugsky-micro and plugsky-lite, plus two API keys and no credit card requirement. It is the same OpenAI-compatible API used by paid plans.
How much do paid plans cost?
Paid self-serve plans are flat monthly, and prices change over time, so check the live pricing page for current figures and limits. There are no per-token charges or overage fees on self-serve plans.
Is there a trial?
Yes. A 14-day full-access trial lets you run production-like workloads across the catalogue before choosing a paid plan. The free plan remains available for lighter evaluation.
Do I pay per model?
No. Self-serve plans cover the catalogue under one flat plan, with 30+ models available through the same API. Enterprise agreements can specify a negotiated catalogue and capacity.
What happens when I exceed fair-use limits?
Requests are rate-limited with 429 backpressure rather than billed. No overage fee applies on self-serve plans; if sustained usage needs guarantees, enterprise committed capacity is the next step.
Can I use the free models in production?
They are best for prototyping and lightweight production use where their capabilities fit. Move to a paid plan when you need frontier models, higher limits or headroom for traffic peaks.
Does Plugsky require a long contract?
No for self-serve plans, which are monthly and flat-rate. Enterprise agreements are negotiated and typically annual, covering committed capacity, SLA and deployment options.
Where can I see the current plans and prices?
The live pricing page at /#sec-pricing always reflects current plans. Articles deliberately avoid hard-coded prices because they change.