Playground
Test prompts, model settings, and response cost before moving a model into production.
The Playground is a chat sandbox for trying out any model available to your organization. Use it to sanity-check a model's answers, tone, and cost before you deploy it or point real traffic at it.
Usage stats bar
Three live counters sit above the chat window and update after every message you send:
| Stat | What it means |
|---|---|
| Tokens (M) | Total tokens consumed by the conversation so far, in millions. |
| Time | Response latency for the last message, in milliseconds. |
| Total Cost | Running dollar cost of the conversation based on the selected model's pricing. |
Chat window
Your messages appear on the right; the model's replies appear on the left, labeled Assistant. For reasoning-style models, the assistant's full chain of thought is shown inline above its final answer, so you can see how it arrived at a response — useful for judging whether a model is a good fit before you deploy it.
Configuration panel
The right-hand panel controls how the next message is generated:
| Field | What it does |
|---|---|
| Model | Chooses which model in your catalog answers the prompt (e.g. Qwen/Qwen3-0.6B). |
| Temperature | Controls response randomness/creativity. Lower is more deterministic, higher is more varied. |
| Max tokens | Caps how long a single response can be. |
Export & clear
Export chat downloads the current conversation; Clear chat wipes it and resets the usage counters, so you can start a clean test.