Skip to main content
Open the model directory to compare models, capabilities, and prices. Copy a model’s full ID into your request:
The prefix identifies the model’s creator. The gateway manages its serving routes, so the provider in your request log may differ from that prefix.

​
Choose for your task

  1. Match the output. Text, images, speech, transcription, embeddings, reranking, and evaluation use different endpoints.
  2. Check inputs and tools. Confirm the model supports your media and function-calling needs.
  3. Compare context and price. Leave room for the conversation and output, then try representative inputs before switching production traffic.
A bigger or more expensive model is not always the best fit. Measure answer quality, latency, and cost for your own application.

​
Discover models through the API

The live catalog is public and is the authority for model IDs and current prices:
Send your API key to list the catalog as your workspace sees it, including any models enabled only for your workspace:
Model availability also depends on serving-route configuration and, for some models, a completed credit purchase. See errors.

​
Available models

Use the live directory for current prices, promotions, and availability. This generated reference is a snapshot of base catalog rates; it does not include every billing unit or pricing tier.
The table shows base rates. google/gemini-3.6-flash, google/gemini-3.7-flash, and google/gemini-3.8-flash have promotional rates of $0.75 input / $3.75 output per million tokens through December 31, 2026 (UTC).google/gemini-robotics-er-2-preview has a promotional rate of $1.00 input / $5.00 output per million tokens, and google/gemini-3.8-flash-tts of $0.50 input / $9.00 output, both through December 31, 2026 (UTC).mistral/mistral-large-4 has Mistral’s launch rates of $0.68 input / $2.09 output per million tokens through October 19, 2026 (UTC).The zai/glm-5.3-flash promotion of $0.075 input / $0.25 output per million tokens ended September 9, 2026 at 16:00 UTC; its base rates now apply. GET /v1/models returns the effective rates used for billing.
For audio units and examples, see transcription and text to speech. For image-output models, use image generation.
meta/muse-spark-1.3-contributor is Meta’s contributor tier. It is the same model as meta/muse-spark-1.3 — same weights, same 1M-token context, same modalities — and it is cheaper because Meta uses the prompts and completions you send it to improve their products. The standard meta/muse-spark-1.3 is not used for training. Nothing else in the gateway routes between the two: the model id you send is the tier you get, and a request never falls back from one to the other. Send anything confidential to the standard id.

​
Retirement

An announced retirement appears in the model’s deprecation block with retire_at and, when available, replaced_by. The model continues serving until withdrawn. Use the live catalog rather than treating the announced date alone as proof that it has stopped. Once withdrawn, the old ID returns 404; the error can name its replacement. Test the replacement with your application’s inputs before switching.

​
Credits

Usage comes from the key’s workspace. 1 credit = $0.000001. Billing and credits explains fees, usage units, balance checks, and BYOK charges.

​
How a request is priced

Read live rates from GET /v1/models and actual charges from request logs. See the cost calculation for token classes and other billing units.

​
Topping up

Choose add credits in the console, or create a hosted invoice. The minimum purchase is $5; the 5% platform fee and applicable tax are added to the credit amount. Speed tiers offer Fast for anthropic/claude-opus-5-5 and Ultrafast for selected OpenAI models. The catalog lists standard rates; premium rates and model support are documented on that page.