All posts

Mistral OCR 4, desktop apps with Deno and computer use in Gemini 3.5 Flash

Curated by Dexpert, with the help of AI.

// Share this post

dexpert.com.br/en/blog/2026-06-29-mistral-ocr-4-deno-desktop-gemini-computer-use/

Document automation, a new path to desktop apps with web technology and agents that operate browsers, phones and desktops.

Image: Mistral AI

Mistral OCR 4 reads documents in 170 languages and runs in a single container

Mistral launched OCR 4, which besides the text returns the position of each passage on the page, the block type, such as title, table, equation or signature, and a confidence score. The model covers 170 languages and, according to the company, independent annotators preferred it in an average of 72% of comparisons with other systems. It can be used through the API or hosted in a single container, keeping documents in the customer's environment.

Opportunity for your business

Processes with invoices, contracts, forms and reports can be automated more precisely, and confidence scores help send only doubtful cases to human review. Self-hosting serves those with privacy and data protection requirements.

How Dexpert helps: Digital Transformation
Image: Deno Docs

Deno now turns web projects into cross-platform desktop apps

deno desktop, available from Deno 2.9, turns a project, from a single TypeScript file to a Next.js app, into a self-contained desktop application for macOS, Windows and Linux. By default it uses the operating system's own webview, which produces small binaries, with the option to bundle Chromium for identical rendering across platforms. It also detects frameworks such as Next.js, Astro and Nuxt, cross-compiles for all three systems from one machine and includes binary-diff auto-updates.

Opportunity for your business

Companies that need a desktop app, such as internal systems or tools that work offline, can reuse the web code they already have. It is an alternative to Electron with smaller packages and built-in updates.

How Dexpert helps: Enterprise Apps & SaaS

Sources

Image: Google

Gemini 3.5 Flash gets native computer use for agents on browser, mobile and desktop

Computer use, previously available only in a separate Gemini 2.5 model, is now a native tool in Gemini 3.5 Flash. With it, developers can build agents that see the screen, reason and act across browsers, mobile apps and desktop, in tasks such as continuous software testing and work in professional applications. Google also offers optional enterprise safeguards, such as requiring user confirmation for sensitive or irreversible actions.

Opportunity for your business

UI testing, routines in systems without APIs and repetitive tasks across several applications come within reach of agents. To use it safely, combine human confirmation, isolated environments and restricted permissions.

How Dexpert helps: Artificial Intelligence

// Share this post

dexpert.com.br/en/blog/2026-06-29-mistral-ocr-4-deno-desktop-gemini-computer-use/

Curated by Dexpert, with the help of AI.

Want to put one of these to work in your company?

Dexpert builds custom systems, apps, integrations and AI solutions. Tell us what you need and get an estimate, or take the free digital maturity assessment.