Models

What is plugsky-tiny and when should you use it?

plugsky-tiny is the small, fast profile in Plugsky's 30+ model catalogue, intended for high-volume, low-complexity calls where responsiveness matters: short chat, tagging, routing decisions and simple extraction. It supports streaming, function calling and JSON mode on the OpenAI-compatible API with a 128K-class context. Choose it for cheap first hops inside a routing chain, and escalate to a stronger tier when a validator fails or the task needs reasoning.

Key facts

Model classSmall, fast profile in the Plugsky catalogue
Best forShort chat, tagging, routing decisions and simple extraction
Context class128K-class window; live limits are published per model
CapabilitiesStreaming, function calling and JSON mode
Pricing tierPaid-plan model; free plan covers plugsky-micro and plugsky-lite
APIOpenAI-compatible /v1/chat/completions — no SDK changes
ResilienceBackup upstream plus same-profile fallback peers
Product statusLive

TL;DR

  • A small, fast tier for high-volume, low-complexity calls.
  • Ideal first hop in a routing or escalation chain.
  • Function calling and JSON mode keep pipelines simple.
  • Do not stretch it to reasoning-heavy work — escalate instead.
  • Compare with the free tiers before paying for the same job.

How it works, step by step

  1. Read the live model card at /models for window, limits and feature flags.
  2. Pick tasks with simple outputs: tags, routes, short replies, small fields.
  3. Benchmark plugsky-tiny against plugsky-lite and plugsky-micro on those tasks.
  4. Add a schema or rule validator to catch weak answers automatically.
  5. Escalate failures to a stronger tier with one capped retry.
  6. Track escalation rate and latency percentiles as volume grows.
1Read the live modelcard at /models forwindow, limits and2Pick tasks withsimple outputs:tags, routes, short3Benchmarkplugsky-tinyagainst4Add a schema orrule validator tocatch weak answers5Escalate failuresto a stronger tierwith one capped6Track escalationrate and latencypercentiles as

Try it yourself

Open the API latency tester →

What plugsky-tiny is

plugsky-tiny is the small, fast profile in the Plugsky catalogue. It targets high-volume calls where the job is simple and the answer must arrive quickly: tagging, routing, short replies, simple extraction and utility prompts. It keeps the standard OpenAI-compatible surface — streaming, function calling and JSON mode — and is listed as 128K-class on the live model card.

Check /models for the current engine, limits and routing, and compare it with the free tiers before paying for the same simple job.

When to choose it

Choose plugsky-tiny as a first hop in a chain. Because it is small, it is well suited to triage stages: classify the request, decide the route, answer trivial questions directly, and hand anything harder to a larger model. That pattern keeps median latency and cost down while protecting quality with an escalation path.

It is a poor fit for multi-step reasoning, ambiguous instructions or long synthesis. For those, start with plugsky-minimax or plugsky-coder-fast for coding, or escalate upward.

Small-model trade-offs

Speed comes with narrower competence. The failure modes to expect:

  • Confident answers to ambiguous prompts — keep instructions explicit.
  • Occasional format drift — constrain with JSON mode and validate.
  • Weak performance on anything multi-step — route rather than retry.

Measure latency as percentiles, not a single average, and watch the escalation rate; if most requests escalate, tiny is the wrong first hop for that workload. Self-serve plans are flat monthly with fair-use usage — see the live pricing page for current plans.

How to switch to plugsky-tiny

Switching is one model name on the same endpoint: {"model": "plugsky-tiny", "messages": [{"role": "user", "content": "Classify this message as billing, technical or other."}]}.

Run it on a sample of real traffic against your current default, compare pass rate and latency, then adopt it for triage flows with a stronger model behind it for escalations.

Honest comparison

Dimensionplugsky-tinyFree models (micro, lite)Larger workhorse tiers
Best fitTriage, tagging and short repliesGeneral free experimentationReasoning and synthesis
SpeedFast tierFast tierModerate
Quality ceilingGood for simple tasksGood for simple tasksHigher
Context class128K-class128K-class128K-class
Use asCheap first hopZero-cost baselineEscalation target
PlanPaidFree, no cardPaid

Frequently asked questions

Is plugsky-tiny free?

No — it is a paid-plan model. The free plan covers plugsky-micro and plugsky-lite with no card required, and a 14-day full-access trial lets you evaluate paid tiers first.

What is plugsky-tiny best at?

High-volume, low-complexity work: tagging, routing, short chat replies and simple extraction where responsiveness matters more than depth.

How does it differ from the free models?

Feature-wise it is similar in shape; the difference is positioning and capacity. Benchmark it against plugsky-lite and plugsky-micro on your tasks before paying for the same job.

What context window does it have?

It is 128K-class today, but the live window and output limit are published per model on the catalogue.

Can it be used for agents?

It can handle simple tool-argument generation, but keep complex planning on a stronger tier. Cap tool loops and validate arguments server-side.

When should I escalate?

Whenever a validator fails or the request needs multi-step reasoning. Escalate once, then hand over to a human path if the stronger model also fails.

How is pricing structured?

Self-serve plans are flat monthly with fair-use usage rather than per-token billing. See the live pricing page for current plans.