Playground & Model Catalog
Try models in the browser, compare pricing, capability and measured performance, and pick the right model.
Choose and validate your model in the browser before writing code — it is by far the fastest order of operations.
Playground
Menu location: Chat → Playground in the left sidebar. Talk to models directly, no code required.
Steps:
- Pick a supplier (the route determines which models and prices are available);
- Pick a model;
- Send a message; multi-turn conversation is supported;
- Some models accept image input or generate images — follow the on-page hints.
Switching supplier refreshes the model list automatically, showing only models that route genuinely serves. If your currently selected model is unavailable on the new route, the system switches to the first available one.
This helps you avoid "no available channel" errors — anything selectable in the Playground is callable via the API on the same route.
Playground calls are billed exactly like API calls — deducted from your account balance and recorded in usage logs. It is not a free sandbox.
The Playground is best for:
- comparing several models on your own real prompts;
- verifying that a supplier route works;
- checking a model's output style and instruction-following.
Model catalog
The entry point is in the top navigation (usually labelled "Pricing" or "Models") and does not require signing in.
It lists every model available on the site:
| Information | Description |
|---|---|
| Model name | The exact value for the model field — copy it, do not type it |
| Price | Input / output unit price, or per-call price |
| Capability tags | Function calling, vision input, web search and so on |
| Performance | Some models show recent measured throughput and first-token latency |
Click a model for its detail page with description, use cases and code samples.
Prices are listed per supplier. The same model may cost differently on different routes, so read the row matching the route you will actually use.
Always copy the model name
A misspelled model name is the number-one cause of no available channel errors. Case, hyphens and version suffixes all matter. Make copying from the catalog a habit.
Rankings
The top navigation may also include Rankings, ordering models by measured data — handy for answering "which model performs best at this price point".
Recommended selection flow
- Model catalog: shortlist two or three candidates on price and capability;
- Playground: test each on your real prompts, comparing quality and speed;
- Once decided: create a key under API Keys and integrate per the API Reference;
- After launch: watch real costs in usage logs and revisit step 1 if needed.
Do not choose on unit price alone. A cheap model with weaker output often needs more turns and longer prompts, so the total cost can be higher. Running a few real samples in the Playground beats reasoning from the price list.