Firefox Nightly Smart Window exposes real AI models and adds custom local LLM support

Firefox Nightly now reveals the actual AI models behind Smart Window and introduces an option to connect a custom local language model. The change moves Smart Window from interface structure to visible AI processing inside the browser.

Firefox’s Smart Window first appeared as a separate browsing mode with memory controls and model selection. The latest Nightly version replaces generic labels with identifiable model backends tied to each assistant option.

Smart Window now lists real assistant models

The model choices appear under Settings > AI Controls >Personalize Smart Window> Assistant model in Firefox Nightly.

Smart Window settings map response styles to specific models:

• Fast uses gemini-flash-lite for quick answers.
• Flexible uses Qwen3-235B-A22B-throughput for broader everyday responses.
• Personal uses gpt-oss-120b for more tailored answers.

These model names point to different external language model providers connected to Smart Window in Nightly. Concrete model names replace placeholder descriptions and confirm that Smart Window connects to external language model providers.

Smart Window now supports custom local LLM connections in settings

A new “Use your own LLM” option lets Smart Window connect to a user-defined model.

Firefox provides fields for:

• Model name
• Model endpoint
• API key or authentication token

Smart Window can connect to locally hosted or self-managed models, something most mainstream browsers do not expose. Firefox also displays a warning that custom models may not behave as expected, which keeps the feature limited to Nightly.

Initial Smart Window versions introduced a separate window mode instead of a sidebar, along with memory and personalization controls and model selection during first setup. Visible model backends and custom configuration now show working AI integration, not only interface structure.

Smart Window is still limited to Nightly, but real model connections and local LLM configuration are now visible. Firefox now handles AI processing inside the browser instead of presenting AI only as an interface feature.