Data layout
The generated public dataset is data/all.json. It is a JSON object keyed by provider ID.
{
"openai": {
"id": "openai",
"name": "OpenAI",
"website": "https://openai.com",
"apiBaseUrl": "https://api.openai.com/v1",
"models": {
"gpt-5.5": {
"id": "gpt-5.5",
"pricing": { "input": 1.25, "output": 10 },
"limit": { "context": 128000 },
"features": { "tool_call": true }
}
}
}
}
There is also data/all.min.json, which has the same shape but is minified.
Provider entries
Each provider has:
id: stable provider ID.name: display name.website: provider website when known.apiBaseUrl: API base URL when known.aiSdk: optional Vercel AI SDK package/env metadata.models: object keyed by model ID.
Model entries
Each model has an id and may include:
namepricing: per-million token pricing fields likeinput,output,reasoning,cache_read, and audio pricing.limit:context,input, andoutputtoken limits.modalities: supported input and output modalities.features: booleans for attachments, reasoning, tool calls, structured output, and temperature.knowledge_cutoff,release_date,last_updated: Unix timestamp strings.open_weights: whether the model weights are open.
Source files
The updater stores source data as TOML under data/providers, then generates all.json from it.
data/providers/
openai/
index.toml
logo-black.svg
logo-raw.svg
models/
gpt-5-5/
index.toml
metadata.toml
Do not read the raw TOML files in applications. Use all.json, all.min.json, or the npm package.