Explore the catalogue
How to find the right model quickly, by category, by capability, by price, or by comparing candidates side by side.
Last updated 2026-07-24
The model directory is the workbench. Every filter you apply is written to the URL, so a filtered view is a link you can send to a colleague or paste into a ticket.
Browse by category
Twelve categories organise the catalogue by what a model is for: language, reasoning, coding, vision, image, video, audio, speech, embeddings, agents, search and research. Each category landing page explains the category before it lists models, including how to choose within it.
Filter precisely
Beyond category, the directory filters on provider, capability (tool calling, JSON mode, streaming, vision input, long context, open weights and more), context window band, price band, country of origin and open-weights status.
| Facet | Example | Useful when |
|---|---|---|
| Context window | 129k – 1M | Long documents or whole codebases |
| Price band | Under $0.50 / 1M | High-volume or batch workloads |
| Capability | Tool calling + JSON | Agentic or structured pipelines |
| Open weights | On | Self-hosting or licence requirements |
Compare candidates
Add up to four models to the compare tray from any card, then open the comparison workspace. Pricing, context, latency, capabilities and benchmark groups line up row by row, with the best value in each row marked. The comparison lives in the URL too, a shareable artefact rather than a session.
Tip: Press ⌘K (Ctrl+K on Windows) anywhere on the site to search models, providers, categories and docs from one field.