Skip to content

Models

Every agent runs a catalogue model. The Models tab is the Flowstate-maintained catalogue of AI models: what they are, what they can do, and what they cost. It lives at Resourcing → Agentic workforce → Models (/plan/main/resourcing/agents/models).

Why a catalogue

An agent stores a reference to a model, never a copy of its price. When a catalogue price changes, the change flows through every agent, forecast, and cost line at once. Agent costs never drift out of sync with what the model actually costs.

If a model has no catalogue price yet, agents running it show Awaiting price data in their cost column until the price lands.

What each model carries

The list shows one row per model, newest first. Each row carries:

FieldWhat it shows
ModelThe model name, with the lab that makes it
ProvidersThe providers that serve the model
ContextThe context window, in tokens
TierThe capability tier
Input /1MThe cheapest published input price, in USD per million tokens
Output /1MThe cheapest published output price, in USD per million tokens
UpdatedWhen the entry last changed

Click a row to open the model's detail drawer. It adds the model family, release date, knowledge cutoff, max output, input and output modalities, capability badges (reasoning, tool calling, attachments), whether the weights are open or proprietary, and a full pricing table per serving provider, including cache read and cache write rates. Catalogue data and pricing are sourced from the public models.dev index.

Finding a model

  • Search by model, lab, or provider name.
  • Filter by tier, provider, modality (text, image, video, audio, embeddings), or a minimum or maximum context window.
  • Sort by recency, name, context window, provider count, or blended cost per million tokens.

Filter and sort state is mirrored to the URL, so you can share a filtered view with a colleague.

What you use it for

  • Picking a model when you create or edit an agent. The model field in the agent drawer draws from this catalogue.
  • Comparing tiers and prices before committing an agent to an expensive model. The deploy drawer's recommended config suggests the cheapest tier that clears the agent's observed demand.
  • Reading per-model breakdowns. Agent insights and spend reconciliation break usage down by model, using these catalogue entries.

For how model prices turn into agent cost lines, see The AI cost model.

Flowstate Documentation