Report
Mozilla's Firefox Smart Window beta runs on Mistral models, starting in France and North America
Mistral and Mozilla have announced a partnership that puts Mistral models inside Firefox Smart Window, described as Mozilla's AI browsing assistant in beta and free to use. The announcement states capabilities and rollout regions, but offers no benchmarks, no named models, and no comparison with what Firefox users had before — so the reader question is only partly answered.
Mozilla and Mistral are partnering to power Firefox Smart Window with Mistral models, according to Mistral's announcement. Smart Window is described as Mozilla's AI browsing assistant, currently a beta, running in the browser. It is available in France and North America, with the United Kingdom and Germany expected to follow later this year. The announcement frames the work as bringing "open, private and multilingual" AI to where people already browse, and the page marks the application as free.
Three behaviours are described for the assistant: helping users make sense of complex searches, remembering information they clicked away from, and sourcing information based on their browser tabs. Those are capability descriptions rather than results. The announcement does not name which Mistral models are used, does not report latency, accuracy or cost, and does not say where processing happens or what leaves the browser — a gap worth noting given that privacy is one of the three framing words.
The question this desk asks of any change is what it does that the previous option could not. On this evidence the answer is incomplete: the source does not describe what Firefox users had before, and it does not compare the assistant against any other tool. What the announcement does claim as distinguishing is that the AI is open, private and multilingual, that it lives inside the browser, and that it is free. Tab-grounded sourcing is the most concrete of the described capabilities, because it implies the assistant reads the tabs you have open rather than only the page in front of you. Treat all of that as claims to test, not findings.
A bounded way to check it: if you are in France or North America, spend a week running your normal tab-heavy research and ask one narrow question each time — did the tab-based sourcing point at the tabs you actually had open, and did it work in the languages you read? That is a small, repeatable test of the stated capability. Readers elsewhere have nothing to install yet; for them the thing to watch is real availability, not the "expected" timing for the United Kingdom and Germany.
Our reading
For a desk focused on practical, everyday AI work, the notable part is placement rather than model capability: an assistant that reads your open tabs sits closer to real research workflows than a separate chat window, and free in-browser access lowers the cost of trying it. But the announcement is a partnership statement, not an evaluation, so readers in the launch regions should treat openness,…
What to do or watch
If you are in France or North America, run a short personal test of tab-based sourcing in the languages you actually work in before relying on it; otherwise, watch for confirmed availability in your region rather than the announced expectation, and for any later detail on models, data handling and cost.
Source details and supporting facts
Each line is stated by the page named above it.
Stated by mistral.ai
- Mistral and Mozilla are partnering to power Firefox Smart Window with Mistral models.
- Firefox Smart Window (beta) is Mozilla's AI browsing assistant, powered by Mistral models.
- Firefox Smart Window is available in France and North America, with the United Kingdom and Germany expected to follow later this year.
- Firefox Smart Window helps users make sense of complex searches, remember information they clicked away from, and source information based on their browser tabs.
Sources
- Mistral AI newsText stored 16 September 2026
How this story was checked. Written from the 1 page listed above, stored 16 September 2026; claims checked against that stored text on 16 September 2026.
What that means
- 4 of 5 reported statements were confirmed against the page that carries them; the rest were removed rather than published.
- Figures in the text were required to appear in the stored source text: yes. Identifiers: yes.
- The check reads stored text only: no claim rests on a fresh look that did not happen.
- Where the reporting was silent, the text says so instead of filling the gap.