Manus gets a Max upgrade and mobile app skills

The hardest part of working with autonomous AI agents has always been the babysitting. You give them a prompt, watch them open twenty tabs, and then spend the next hour fixing their broken code.

Manus wants to end that cycle. The platform just launched Manus 1.6, a major release aimed squarely at cutting down human intervention.

The update is headlined by Manus 1.6 Max, a new flagship agent architecture designed for complex planning and one-shot task completion. In double-blind testing, Manus claims user satisfaction jumped over 19.2% thanks to better accuracy and more dependable tool use.

Every sub-agent in its Wide Research feature now runs on this Max architecture for parallel data gathering. The model has also been heavily tuned for spreadsheet work, taking on financial modeling and automated reports that typically break traditional LLMs.

Beyond logic upgrades, Manus is expanding where its code lives. For the first time, users can generate mobile applications end-to-end directly from plain text prompts, moving well past standard web sandboxes.

There is also a new Design View interface. Instead of rerolling an entire image prompt when something looks slightly off, users get an interactive canvas with point-and-click local editing, multi-image compositing, and the ability to cleanly modify in-image text.

The entire 1.6 update is available now to all users directly in the task launcher.

Why it matters

We think this release signals a clear shift in the agent race: the honeymoon phase of flashy browser demos is officially over, and the real game is now output reliability.

For solo founders and operators, building mobile apps alongside internal web dashboards from a single chat prompt removes a massive technical barrier. If you regularly build quick internal tools or wrestle with financial models, it is definitely worth testing Max today to see if it actually spares you the debugging headache.

That said, full-stack mobile generation is a notoriously difficult promise. Web apps are easy to patch on the fly, but native mobile environments involve strict permissions, platform-specific quirks, and brittle dependency trees. If the agent delivers spaghetti code that breaks outside the sandbox, the cleanup could easily eat up whatever time the generation saved.