Every raia agent uses a large language model (LLM). raia cX lets your organization decide which available models your team can use, while each agent uses one model from that approved list. New organizations begin with the active model selection enabled, so setup usually means keeping the models that fit your needs and removing the ones that do not.
In this section
- Understand how organization-level model availability affects each agent.
- Add, archive, restore, and remove models from your organization's approved list.
- Choose a model for an agent and apply the settings that model supports.
- Respond safely when a model is in use, unavailable, or retired.
- Know which organization and agent roles can view or change model settings.
Model providers
raia offers models from the major frontier providers alongside leading open-source families. The models available in your organization are managed by raia, and your organization's admins choose which of them are active for your team.
OpenAI
GPT-5, o-series
Anthropic
Claude 4 family
Gemini 2.5
xAI
Grok
Mistral AI
Mistral Large, Magistral
DeepSeek
DeepSeek-V3, R1
Qwen
Qwen3
Meta
Llama 4
Open-source model families
Open-weight models give you strong alternatives with transparent lineage, and can be a good fit when cost, latency, or deployment flexibility matter. Availability in raia follows the same organization-level model controls as every other model.
Llama
Open-weight models from Meta
Gemma
Lightweight open models from Google
Qwen
Open-weight models from Alibaba
DeepSeek
Open reasoning models
Mistral
Open-weight releases, including Mixtral
Ollama
Run open models on your own hardware
What's inside
- 01
How model availability works
raia maintains the conversation models available to customers. An Organization Owner or Organization Admin chooses which of those models are active for the organization. Anyone who can edit an agent then selects one active organization model for that agent. Only active organization models appear in the agent's model picker.
- 02
Manage your organization's models
Open Settings, then LLM Models. Use Select a model to search or filter available models by provider, then choose Add Model. Use the status control to archive a model temporarily or restore it later. Removing a model takes it off your organization's list entirely; archiving keeps it listed but prevents new assignments.
- 03
Protect agents that already use a model
Before a model can be archived or removed, raia checks whether any agents still use it. If the model is in use, the page lists the affected agents. Reassign those agents first, then archive or remove the model. This safeguard helps prevent an accidental interruption to a live agent.
- 04
Choose a model for an agent
Open the agent, go to Train, select Advanced Settings, and find LLM Configuration. Choose a model from your organization's active list and select Apply. The picker can be searched and filtered by provider, making it easier to compare the options approved for your team.
- 05
Configure model-specific options
The settings shown depend on what the selected model supports. Options may include reasoning effort and mode, temperature, Top-P, file search, code interpreter, structured JSON output, and a default error message. Settings that do not apply to the selected model are not shown.
- 06
Handle an unavailable model
If an agent's saved model is no longer active, its Advanced Settings show that the saved model is unavailable. Select an active replacement and apply the change promptly. There is no automatic fallback. An agent continues working as expected only while its selected model remains available to your organization.
- 07
Choose the right model
Match the model to the agent's work. Models designed for deeper reasoning can suit complex, multi-step tasks, while lighter models can be faster and more economical for straightforward conversations. Test realistic scenarios before changing a live agent. Your organization and per-agent usage limits continue to apply regardless of the model selected.
At a glance
| Action | Who can do it | What to know |
|---|---|---|
| View organization models | Any organization member | Understanding which models are available to the team. |
| Manage organization models | Organization Owner or Organization Admin | Adding, archiving, restoring, or removing approved models. |
| Set an agent's model | Agent Owner, Agent Admin, or Agent Editor; Organization Owner, Organization Admin, or Organization Manager | Selecting and tuning the model for an agent you can edit. |
Before changing a live agent
Confirm that the replacement model supports the settings and capabilities the agent needs, then test representative conversations before applying the change in production.