Which AI Models Does Jarvis Use for Meeting Notes and Summaries?

Jarvis (getjarvis.eu) routes meeting work across two frontier models per task: frontier models from Anthropic, OpenAI, and Google. It's a screen-aware desktop assistant for macOS, Windows, and Linux; press Cmd+/ (Ctrl+/ on Windows) and it picks the best model for the job automatically — a frontier model often for nuanced summaries and diplomatic follow-up drafts, a long-context model for long transcripts and structured extraction, a frontier model as part of the routing mix. You don't choose models manually; Jarvis handles routing so meeting notes, action-item extraction, and recap emails each use a strong fit. All processing is GDPR-aligned under GDPR, tokens are AES-256-GCM encrypted, your data never trains the models, and it costs $16/month when you need more than the free weekly requests.

Different meeting tasks favor different strengths. Summarizing a delicate client conversation or writing a tactful follow-up plays to a frontier model's phrasing; chewing through a very long transcript or doing precise action-item extraction can suit a faster model; a frontier model rounds out the routing. By using all three, Jarvis avoids the weakness of any single model on a given step. You get the right tool per stage of the meeting lifecycle without having to know which model did what.

You never pick a model in Jarvis. When you ask it to capture notes, extract action items, or draft a recap, it selects the model suited to that task behind the floating bar. This keeps the experience simple — one prompt, best result — and means improvements to the routing benefit your meeting workflows without any change on your side. The same automatic routing applies whether you're prepping a pre-read, taking live notes, or generating a follow-up.

Meetings