Cerebras hosts open models on its own inference hardware. The default Edesy runtime supports it with gpt-oss-120b as the default model. Cerebras is not yet in the dashboard model picker, so set it on an agent through the API or ask support to switch the agent for you.
At a glance#
| Item | Value |
|---|---|
| Provider ID | cerebras |
| Runtime | Default runtime |
| Default model | gpt-oss-120b |
| Dashboard picker | Not listed yet. Set it through the API or ask support |
| Your own key (BYOK) | Supported |
Note: the model ID differs from Groq's for the same model. Cerebras uses gpt-oss-120b; Groq uses openai/gpt-oss-120b.
Set it up through the API#
curl -X PATCH https://voice-agent.edesy.in/api/v1/agents/AGENT_ID \
-H "Authorization: Bearer $EDESY_API_KEY" \
-H "Content-Type: application/json" \
-d '{ "llmProvider": "cerebras", "llmModel": "gpt-oss-120b" }'
Your API key needs the agents:write scope. See authentication and scopes.
FAQs#
Why is Cerebras not in the dashboard? The runtime supports it, but the model picker does not list it yet. Agents set to Cerebras through the API run normally.
Will editing the agent in the dashboard reset the model? Check the model field after saving in the dashboard. If it changed, set it again through the API or ask support.
Does Cerebras support function calling? Yes, through the OpenAI-compatible API. Test your tools before going live.