Models and providers
Choose a provider and model: your own API key, a local model, or Lamied-hosted models on the Team plan.
Choose your model first
Lamied defaults to Nemotron 3 Super through its hosted service. Lamied Ultra and Lamied Team subscribers sign in without adding a provider key. You can also use your own API key or a local model. Existing saved model choices are preserved. Change your choice any time with /model.
Supported providers
| Provider | What you need | Models |
|---|---|---|
| Nemotron (NVIDIA) | An NVIDIA API key | One fixed model |
| OpenAI | An OpenAI API key | Choose from your account's models |
| Claude (Anthropic) | An Anthropic API key | Choose from your account's models |
| Gemini | A Google AI (Gemini) API key | Choose from your account's models |
| Groq | A Groq API key | Choose from your account's models |
| Mistral | A Mistral API key | Choose from your account's models; also powers Codestral autocomplete |
| AddisAI | An AddisAI API key | One fixed model |
| Local Model | A model server on your computer (no key) | Whatever your server provides |
| Lamied Ultra | A Lamied Ultra or Team subscription (no provider key) | Amazon Bedrock models |
Connect a provider with your API key
- Create an API key in your provider's dashboard (OpenAI, Anthropic, Google AI Studio, Groq, Mistral…).
- In Lamied, open the AI panel (Ctrl+J) and click ⚙ at the top, or type
/settings. - Choose your provider in the Provider menu.
- Paste the key into API Key and click Save. The status line confirms the key is saved.
- Type
/model, pick the provider, then pick the model you want. The chat confirms it, for example Switched to Claude (Anthropic) · claude-…
You can save keys for several providers and switch between them any time with /model. To delete a key, choose its provider in settings and click Remove key.
How keys are stored
- Keys are encrypted with your operating system's credential store (the keychain on macOS, the credential vault on Windows, the Secret Service keyring on Linux) and saved only on your computer.
- They are never sent to Lamied. Each key is used only for requests to its own provider.
- If your system has no secure credential store, Lamied refuses to save the key rather than store it unencrypted. On Linux, install and unlock a keyring such as GNOME Keyring or KWallet.
Use a local model
- Start a model server with an OpenAI-compatible API, such as Ollama, LM Studio or llama.cpp's server.
- In AI settings, choose Local Model.
- Enter the server's Base URL, for example
http://localhost:11434/v1for Ollama, and click Save. No API key is needed. - Choose the model with
/model.
A local model keeps your prompts on your computer. Signing in and checking Pro still need an internet connection, and a small local model may struggle with long, multi-step tasks.
Amazon Bedrock models
Choose Lamied in AI settings, then use /model to select Nemotron 3 Super, Nemotron Nano 3, Llama 4 Maverick 17B or Llama 4 Scout 17B. Team and Ultra both include these models. Ultra also adds Kimi K3, GLM 5, GLM 4.7 and DeepSeek V3.2. No AWS account or Bedrock key is needed.
Llama 4 tool-enabled turns appear after the complete response arrives; ordinary text turns stream. Use the Plans button in the AI panel to upgrade or renew, including when you already have Pro.
Use Lamied Ultra or Team
- Sign in with your subscription account (or the email your Team owner added).
- In AI settings choose Lamied Ultra. No provider key is required.
- Use
/modelto choose Gemini 3.5 Flash-Lite, GLM 5.3 Flash, DeepSeek V4.1 Flash, Nemotron 3.5 Super (Pro and above) or Kimi K3, GLM 5.3, Gemini 4 Argon, Gemini 3.8 Flash (Ultra and Team).
Ultra has individual 5-hour, weekly and 30-day usage windows. Team shares larger windows among its 3 members. The meter above the message box shows use and reset times. When a limit is reached, switch to a personal API key or wait for the reset. Lamied Pro supports personal provider keys and local models. See Billing and Pro.
Chat model and autocomplete model
/model sets the model that answers in the chat. Autocomplete has its own setting, /fim, because it's called far more often and needs to be fast. See Slash commands.