What is the Plugsky LLM API?

Key facts

Compatibility	OpenAI-compatible /v1/chat/completions with streaming, tools and JSON mode
Models	30+ models from free chat to frontier reasoning and multimodal
Free tier	plugsky-micro and plugsky-lite on the free plan, no card required
Routing	Model routing and fallback chains across models
Context	Model-dependent windows, up to 1M tokens on long-context models
Pricing model	Flat monthly self-serve plans with unlimited fair-use usage
Deployment	Cloud, VPC, on-prem and air-gapped options
Product status	Live

TL;DR

• One endpoint, 30+ models, no code rewrite.
• Start free with plugsky-micro and plugsky-lite, then scale to a paid tier.
• Route simple requests to small models and hard ones to frontier models.
• Fallback chains keep requests alive when a model is unavailable.
• Deployment options range from public cloud to air-gapped.

How it works, step by step

1. Create a key on the free plan, with no card required.
2. Point your OpenAI client at https://plugsky.com/v1.
3. Choose a model per task: micro or lite for simple calls, pro for production, frontier for hard reasoning.
4. Enable streaming for interactive applications.
5. Add model routing or a fallback chain for resilience.
6. Monitor usage and move to a higher tier when rate limits bind.

Original data

Try it yourself

Open the model picker →

The models on Plugsky

The catalogue spans several tiers. Free models such as plugsky-micro and plugsky-lite cover fast reasoning, classification and simple chat. Mid-tier models such as plugsky-plus and plugsky-pro handle general chat, content generation, code and function calling.

Above those sit long-context and multimodal models, hard-reasoning models such as plugsky-frontier and plugsky-reasoning, cost-optimised models such as plugsky-deepseek-flash, and embedding models with 1536 or 3072 dimensions. See the live catalogue for the current list and per-model context windows.

Choosing the right model for the task

The right model depends on three factors: quality required, context length and cost ceiling. Simple classification, intent detection and short replies belong on the free models. General chat and function calling belong on a mid-tier model. Book-length or codebase analysis needs a long-context model, and hard reasoning, maths or complex code belongs on the frontier tier.

Multimodal work moves to the vision models, and high-volume cost-sensitive work to the efficient models. Matching each request to the smallest model that can handle it is the single biggest cost lever.

Routing, failover and limits

Use model routing to pick the right model automatically, or configure a fallback chain so a 503 or rate limit on one model shifts traffic to a peer. The API returns 503 with a Retry-After header when a model is unavailable, which makes client-side fallbacks straightforward.

All models are unlimited under fair use on self-serve plans, with rate limits that scale by tier. Fine-tuning on open base models is available on Enterprise contracts and deploys into your tenant.

The same key covers the full 30+ model catalogue with no separate accounts.

Honest comparison

Need	Recommended Plugsky models	Why	Alternative
Free tier	plugsky-micro, plugsky-lite	No card, fast	Local small models
Production chat	plugsky-pro, plugsky-plus	Balanced quality and speed	Mid-tier models elsewhere
Long context	plugsky-max, plugsky-longctx	Windows up to 1M tokens	Chunking plus RAG
Hard reasoning	plugsky-frontier, plugsky-reasoning	Strongest reasoning tier	Multiple vendors
Multimodal	plugsky-vision, plugsky-ultra	Image and document understanding	Separate vision API
High volume	plugsky-deepseek-flash, plugsky-phi	Fast and efficient	Self-hosting at saturation