Groq hosts open-weight language models on fast inference hardware. It is available on the default Edesy runtime and in the dashboard model picker. The default Groq model is openai/gpt-oss-120b. Use Groq when you want an open model for a pipeline agent and quick replies matter more than the largest context window.
At a glance#
| Item | Value |
|---|---|
| Provider ID | groq |
| Runtime | Default runtime |
| Default model | openai/gpt-oss-120b |
| Other models in the dashboard | openai/gpt-oss-20b, llama-3.3-70b-versatile, llama-3.1-8b-instant, qwen/qwen3.6-27b |
| Model list | Synced from Groq's catalog, because Groq retires models often |
| Tool calling | Supported through the OpenAI-compatible API |
| Your own key (BYOK) | Supported |
When to use Groq#
- You want an open model for a prompt-based or flow agent.
- Your agent uses short, fast turns (order status, reminders, qualification).
- You want to compare cost and quality against OpenAI or Gemini.
Set it up#
- Open the agent and go to Model.
- Choose Groq as the LLM provider.
- Keep
openai/gpt-oss-120bunless you have tested another model. - Save and run a few test calls, including calls that use your tools.
{
"llmProvider": "groq",
"llmModel": "openai/gpt-oss-120b"
}
FAQs#
What happens when Groq retires a model?
The dashboard model list is re-synced from Groq's catalog. If your agent uses a retired model, switch it to a current one. The default is openai/gpt-oss-120b.
Does Groq support function calling? Yes. Test your tools with a few calls, since smaller models can be less reliable at calling tools.
Can I use Groq with native audio? No. Groq is a text LLM for the STT + LLM + TTS pipeline.
Can I bring my own Groq key? Yes. See BYOK.