Models, families, and labs
Model specs, family lineage, researcher records, release metadata, and capability flags would be available as JSON.
Data productNot yet built - gauging interest
LLMReference API
Programmatic access to the same model, provider, pricing, benchmark, and freshness data that powers LLMReference. Not built yet - tell us if you would use it.
What it would expose
- Format
- Maintained reference data as JSON
- Count basis
- Public language-model rows, including deprecated entries
2,005
Public model rows
151
Providers
2,495
Routes
180
Benchmarks
Provider routes and pricing
Provider availability, token prices, batch prices, cache pricing, and route-level notes would be available as JSON.
Benchmarks and task signals
Benchmark definitions, tracked scores, task tags, and quality signals would be available as JSON.
Freshness and agent surfaces
Last-researched dates, update signals, and agent or harness catalog entries would be available as JSON.
Why we would build it
Tool builders
Embed current model, price, and provider data without rebuilding the same catalog in every product.
Internal dashboards
Keep engineering and finance views aligned around fresh provider pricing and model availability.
Agent frameworks
Consume benchmark, capability, and context metadata when choosing defaults or routing workloads.
What you can use today
These live pages cover the decision work while the LLMReference API remains future scope.
Ranked third-party LLM APIs
Use the existing API leaderboard when you are choosing a provider or model route today.
Per-model decision pages
Open a model detail page for the current verdict, provider ladder, pricing, and freshness notes.
Head-to-head comparisons
Compare two models directly when you need to settle an implementation choice.
Last refreshed 2026-09-30. Next refresh: weekly.
Freshness refers to LLMReference data on the site, not to a launched API.