Choose and Configure a Model
Use official models, your own API key, or a local model, and distinguish models from agent executors.

Choose a model beside the chat input. Configure it in Settings → Model services.
What is in this category?
| Capability or service | What it does | How to use it |
|---|---|---|
| Text models | Chat, summaries, writing, and coding | Choose one in the chat model selector |
| Models with image input | Understand screenshots, photos, or charts | Choose a compatible model and attach an image |
| Official model service | Use models and allowance available to your account | Sign in and inspect the Nexts AI model list |
| Your own API service | Use your provider account and key | Model Services → Add provider → Fetch models |
| Local model service | Call models running locally, such as through Ollama | Start the service, configure its address, and refresh |
| Voice, image, and video models | Transcribe, converse by voice, or generate media | See Voice & Video |
Check the current selector and service configuration for model names and capabilities. One model may not support every task above.
Choose a setup
| Setup | Requires | Use it when |
|---|---|---|
| Official Nexts AI | Sign-in, model access, and allowance | You want to get started directly |
| Your own API key | Provider account, valid key, and available models | You already use a provider |
| Local model | An installed and running local model service | You want to use your computer's resources |
Models generate and understand content. Agent executors organize tool actions. Selecting a text model does not install or sign into every external agent.
Use an official model
- Sign in and open the model picker.
- Choose a model available to your account.
- Send a simple test message.
- For images or other attachments, select a model supporting that input type.
Availability depends on the current service and plan. Check account access and usage if necessary. Official service credentials are not something you need to enter.
Add your own provider
- Open Settings → Model services and select or add the provider.
- Enter the service URL and API key as requested, then save.
- Refresh the model list and enable the models you need.
- Select one beside the input and test it.
Keep keys in dedicated settings. Incorrect URLs, expired keys, and exhausted provider quotas can all cause failures.
Use a local model
Install and start a supported service such as Ollama, download a model, configure its address, and refresh the list. The model needs sufficient machine resources and the service must remain running.
Configure search and media separately
Web search has a separate provider selection. Transcription and voice settings live in Media services. For images and video, see Canvas. Working chat does not prove every other service is configured.
If something goes wrong
- Empty model list: save, refresh, and check the URL, key, and enabled state.
- Chat works but images fail: choose a model supporting image input.
- A model switch seems ineffective: inspect the current conversation's selection and test a new chat.
- An external agent cannot start: check installation, sign-in, and executor status under Agents.
Cloud models receive the content needed for the task. Even a local model can use tools that contact search providers or external apps.