Key facts
| Model class | Top capability tier in the Plugsky model ladder |
| Best for | Hard reasoning, deep analysis and high-value long-context tasks |
| Context class | 128K-class window; live limits published per model |
| Capabilities | Streaming, function calling, JSON mode, reasoning and long-context |
| Pricing tier | Paid-plan model; free plan covers plugsky-micro and plugsky-lite |
| API | OpenAI-compatible /v1/chat/completions — no SDK changes |
| Resilience | Backup upstream plus same-profile top-tier fallback peers |
| Product status | Live |
TL;DR
- The quality ceiling of the catalogue — use it where a wrong answer is costly.
- Not a default: route it to escalations, evaluations and premium workflows.
- Function calling and JSON mode keep it usable in existing agents.
- Peer failover keeps top-tier capacity available during upstream incidents.
- Benchmark it on your hardest cases with the 14-day full-access trial.
How it works, step by step
- Define which requests genuinely need top-tier quality and which do not.
- Read the live model card at /models for window, output limit and feature flags.
- Build an escalation eval set from your hardest real requests.
- Compare plugsky-frontier with your current best model on that set.
- Route qualifying traffic to frontier and keep a faster model as the default.
- Switch by changing the model name, then watch latency and quality dashboards.
- Review the escalation rules regularly so cost stays aligned with value.
Try it yourself
Open the LLM cost calculator →
What plugsky-frontier is
plugsky-frontier is the top capability tier in the Plugsky catalogue. It is intended for requests where the quality of the answer matters more than the time it takes to produce: complex analysis, high-stakes drafting, difficult reasoning chains and long-context synthesis across many documents. It speaks the same OpenAI-compatible chat surface as the rest of the catalogue, including streaming, function calling and JSON mode.
Because the frontier tier is updated most often, the live model card at /models is the source of truth for context window, output limits and capability flags. Read it before wiring constraints into your application.
When to choose it
Choose plugsky-frontier as an escalation target rather than a default. Good triggers include failed validation on a cheaper model, requests flagged as complex, premium customer tiers, and offline jobs where throughput is irrelevant. It is also the natural model for benchmarking: if your product needs a quality reference point, this is the tier to measure against.
For routine chat, extraction and agent steps, use a faster model such as plugsky-deepseek-flash or the platform default workhorse, and escalate only when needed. See when to use a frontier-tier model for routing patterns.
Cost, quality and latency trade-offs
Top-tier quality carries top-tier latency and cost. The engineering task is not to avoid it but to place it where it earns its keep.
- Gate escalation on validation failures, complexity signals or explicit user intent.
- Send complete, well-structured context; a frontier model fed partial data is wasted spend.
- Cache or store outputs for repeated premium tasks instead of regenerating them.
- Self-serve plans are flat monthly with fair-use usage — see the live pricing page for current plans.
How to switch to plugsky-frontier
Switching is one model name on the OpenAI-compatible endpoint: {"model": "plugsky-frontier", "messages": [{"role": "user", "content": "Analyse these contracts and list the three biggest risks."}], "stream": true}.
Test on your hardest cases, compare against your current top model, then route only the qualifying traffic. Keep a faster model configured for the default path so quality escalations never become the norm.
Honest comparison
| Dimension | plugsky-frontier | Workhorse models (e.g. plugsky-pro) | Small models (e.g. plugsky-lite) |
|---|---|---|---|
| Best fit | Hard reasoning and high-value analysis | Everyday chat, agents and coding | High-volume simple tasks |
| Quality ceiling | Highest in the catalogue | Strong general purpose | Good for narrow, simple tasks |
| Latency profile | Slowest tier | Balanced | Fastest tier |
| Context class | 128K-class | 128K-class | 128K-class |
| Use as | Escalation and benchmark | Default production model | Cheap first hop |
| Failover | Automatic top-tier peer fallback | Automatic peer fallback | Automatic peer fallback |
Frequently asked questions
Is plugsky-frontier free?
No — it is a paid-plan model. The free plan includes two free models, plugsky-micro and plugsky-lite. A 14-day full-access trial lets you evaluate paid tiers first.
What context window does it have?
It is 128K-class today, but the exact live window and output limit are on the model catalogue at /models. Use that page as the source of truth.
Should I use plugsky-frontier for everything?
No. It is the slowest and most expensive tier. Use it for hard, high-value requests and keep a faster model as the default for routine traffic.
Does it support tools and structured output?
Yes — streaming, function calling and JSON mode are part of the OpenAI-compatible surface, so existing tool-using code works unchanged.
How is it different from plugsky-max?
Both are top-tier. plugsky-frontier is the frontier-profile flagship, while plugsky-max is the largest-context generalist in the same ladder tier. Evaluate both on your hardest prompts and route accordingly.
What happens if the frontier upstream has an incident?
Requests retry through a backup upstream and same-profile top-tier peers. Live component health is published on the status page.
How do I control spend if usage grows?
Keep escalation gated and measured. Self-serve plans are flat monthly with fair-use usage rather than per-token billing — see the live pricing page for current plans.
Can we deploy it in a private environment?
Enterprise options include your VPC, on-prem and air-gapped deployments. Ask the team which models are available for each option.