DocsCore ConceptsModel Tiers & Capabilities

Model Tiers & Capabilities

How models are categorized and what capabilities mean.

Models in PeerLM are organized into four tiers based on their pricing and capability level. Tiers determine plan availability — which models you may run as candidates, and which may sit on the judge panel.

Tier Definitions

  • Standard — Smaller, faster models. Available on all plans as candidates and as judges.
  • Advanced — Mid-range models with strong general performance. Available on all plans as candidates and as judges.
  • Premium — Large frontier-class models. Pro, Team, and Enterprise as candidates. As judges: Team and Enterprise, assigned per Monitor.
  • Frontier — Newest cutting-edge models. Enterprise only, as candidates or judges.

Model Capabilities

Each model has a capabilities profile that PeerLM checks before making API calls:

  • Temperature — whether the model accepts a temperature parameter
  • Seed — whether deterministic seeding is supported
  • System prompt — whether the model supports a system message
  • JSON mode — whether structured output is supported
  • Max tokens — the model's maximum output length

If a capability is not supported, PeerLM automatically omits that parameter. You never need to worry about sending unsupported parameters.

Model Sync

PeerLM syncs model versions daily. When a model is updated by its provider, the old version is marked as superseded and a new record is created. Your evaluation results are always tied to the specific model version used at run time.

Browsing Models

Go to Settings > Models to browse all 200+ available models. Filter by provider, tier, or search by name. The page shows the last sync timestamp.