Guide Just Got Smarter Again

We've upgraded Guide's AI to Anthropic's newest model, and the improvement is significant. On OS-World, a standardized benchmark for real-world computer tasks, the new model scores 72.5% vs 61.4% on the previous version. That's an 18% improvement on benchmarks measuring the AI's ability to complete tasks on a Windows PC.

---

What This Means for You

Guide's AI now handles your requests with:

• Sharper reasoning. Better at figuring out the right sequence of steps to get something done, especially for multi-step tasks.
• Fewer failed attempts. Higher benchmark scores translate directly to more commands that work on the first try.
• More reliable context. If you've built up a long conversation or given Guide detailed instructions, it holds onto that context more consistently throughout your session.

Why OS-World Matters

OS-World is an independently developed benchmark that tests AI models on real computer tasks: navigating applications, filling out forms, moving files. It's one of the closest measures we have to how good an AI is at actually using a computer. An 18% improvement on that benchmark is one of the largest single-model jumps we've seen.

Why This Matters for Accessibility

For users who rely on Guide as an assistive technology, a smarter model means fewer frustrating moments where Guide misunderstands or fails. Every improvement to accuracy is a direct improvement to independence. Less time correcting mistakes, more time getting things done.

How to Update

Guide updates automatically in the background while you use it. When the update is ready, you'll be prompted to restart. No manual download needed.

If you'd like to install fresh, you can always grab the latest version from our website.

---

Thank you for using Guide. Every piece of feedback you've shared has helped us know where to push. More updates coming soon.

Andrew Co-Founder, Guide