Skip to content

Models

We serve a curated catalog of open-weight models. The browsable, always current list — with pricing, context lengths and capabilities — lives in the app:

Browse the model catalog →

This page explains how to read that catalog and how to call what’s in it.

The string you pass in "model" is the model’s canonical slug, in provider/model form:

{ "model": "zai/glm-5.2" }

The slug is what GET /api/v1/models returns as id, and what the app shows as Model ID on each model’s page. Short internal IDs also resolve, but the slug is the stable, documented identifier — prefer it.

Every model is one or both of the following. The catalog labels each one.

Serverless — live, pay-per-token, no provisioning. Pass the slug and call it.

Dedicated — the model runs on GPU capacity reserved for you. This is currently arranged with our team rather than self-serve; models that support it show an enquiry option in the app. Per-hour prices shown alongside dedicated models are indicative.

The catalog spans more than chat. Each model carries one or more categories, which determine the endpoint you call it on:

Category Endpoint
llm /chat/completions
vision /chat/completions with image content parts
embedding /embeddings
reranker /rerank or /reranking, depending on the model
image /images/generations
audio /audio/transcriptions, /audio/translations

These are callable today with nothing to provision:

Model Category Called via
zai/glm-5.2 llm /chat/completions
google/gemma-4-26b-a4b-it llm /chat/completions
qwen/qwen3-coder-30b-a3b-instruct llm /chat/completions
meta/llama-3.1-8b-instruct llm /chat/completions
openai/whisper-large-v3-turbo audio /audio/transcriptions
black-forest-labs/flux-1-schnell image /images/generations

The catalog is authoritative — check there for context lengths, capabilities and current pricing, and call GET /api/v1/models for the list your key can actually reach.

Pricing is per model and is published in the catalog, which is always current — we don’t duplicate rates here, so you never have to wonder whether a documented figure is stale.

What’s consistent across every model:

  • Prices are in AUD and quoted excluding GST. GST is applied and shown on your invoices.
  • Text models are priced per 1M tokens, with input and output charged at different rates. Image generation is priced per image, and audio per minute.
  • Where a model supports prompt caching, cached input is charged at a lower rate. Cached tokens are a subset of input tokens, not an additional charge — they are never billed twice.

You can also read the rates that apply to your workspace straight from the API, in the pricing object returned by GET /api/v1/models.

Terminal window
curl https://api.resetdata.ai/api/v1/models \
-H "Authorization: Bearer $RESETDATA_API_KEY"

Returns an OpenAI-style { "object": "list", "data": [...] }. By default it lists only models you can call right now. Each entry includes id, context_length, max_completion_tokens, pricing, architecture, model_categories and supported_parameters.

For a single model — note detail is singular, and the slug is a query parameter rather than a path segment:

Terminal window
curl -i "https://api.resetdata.ai/api/v1/models/detail?slug=zai/glm-5.2" \
-H "Authorization: Bearer $RESETDATA_API_KEY"

Context length, output limits, tool-calling support and accepted sampling parameters all differ between models. See Model parameters — particularly if you are switching a working integration from one model to another.