Models
Every way to connect a model in Atlarix: the managed Core catalog, your own API keys, and local inference.
Atlarix Core (managed models)
Core is the managed model catalog built into Atlarix. Sign in, pick a model from the picker, and start — no API key needed for Core itself.
Core is a small, curated set of current open-weight frontier models rather than an exhaustive list. Atlarix chooses them, pays the providers directly, and swaps them as better ones ship — the set changes without an app update, so what you see in the picker is always the current lineup. If you want a specific model that isn't in Core, add your own key (below) and use it that way.
Where Core requests go: through Atlarix's authenticated proxy, which does not log or retain them, to the provider that serves the model — processed there under that provider's own data policy. Local models and your own API keys never touch the proxy.
Core is pay-as-you-go on every plan: keep a prepaid credit balance (top up from $10) that is metered as you use it. Both Free and Pro pay the same way — you are billed for what a reply actually costs, including any background work like sub-agents or context compaction.
Core models today:
The live model list is loaded from the Core models catalog. Enable JavaScript to see the current models.
Your own API keys
Connect any of the 145+ providers in the live models.dev catalog. Popular providers include OpenAI, Anthropic, Google Gemini, and any OpenAI-compatible endpoint.
Add keys under Settings → AI for the active workspace. Keys are stored in your OS keychain (macOS Keychain, Windows Credential Manager, or the system secret store on Linux) and never sent to Atlarix servers.
Your-key requests go directly to the provider — Atlarix acts as a local client. Estimated cost per reply is shown in the chat for your-key turns.
Local models (Ollama / LM Studio)
Run open-weight models entirely on your own hardware. Connect Ollama, LM Studio, or any local endpoint serving the OpenAI-compatible API.
With local models:
- Nothing leaves your machine — fully offline, fully air-gapped.
- No cost — inference is free on your hardware.
- Works the same as any other model in the picker.
Local inference is free on every plan, including Free.
Switching and mixing models
Pick the active model per chat from the model picker — Core, your-key, or local. Switch anytime without leaving the workspace.
Core and your-key providers coexist: if your Core credit balance runs out mid-task, Atlarix prompts you to top up or switch to another model and keeps your conversation intact.
This flexibility means your AI strategy stays yours. If a vendor changes terms, doubles prices, or goes down — you switch in one click. No migration, no retraining.