Replies come from your own model. llmrun doesn't manage this model.
llmrun doesn't store your chats, keys or settings: everything stays only in this browser. We only collect general web analytics on page usage, never message content.
The OpenAI-compatible URL of your model, usually ending in /v1.
For RunPod, use your RunPod API key.
Auto picks the first model available. You can also type a model name.
Balanced defaults that work across most models.
Higher is more creative.
Limits word choice to the likeliest.
Reasoning models need room to think.
Leave a field empty to use the server's default.
Encourages new topics.
Discourages repeating words.
Same seed, same reply.
Up to 4, separated by commas. The reply ends when one appears.
After the first reply, the model suggests a short title.
Show thinking
Let the model look things up. Tools use your own keys and work only with models that support tool calling.
Provider
Get a free Tavily key
Works without a key; may be rate limited.
Get a Jina key
Download 0 chats and your settings as a file.
Add chats and settings from an export file.
Remove all chats, settings and keys from this browser.