Skales v12.9.35 - Roles
We've once again cut token spending: an ordinary turn now sends about 45% fewer input tokens on desktop and more than half fewer on mobile, a simple greeting 70–80% fewer, and Skales Code and Flow start 40–50% lighter, with the same answers as before. An idle Skales window also does about a third less work.
We've just released Skales v12.9.35, bringing a number of useful refinements to your desktop AI experience. This update focuses on improving how you manage and interact with your AI models, making daily use smoother and more predictable.
Roles and Model Selection
The "Roles" feature has been updated to function as a true switch, allowing you to easily turn it on and off. When active, it clearly indicates which model each role is using. You can now assign specific models to your sub-agents in both chat and Code, choosing from a live list of available models. This means you can use a more economical model for background tasks while your main conversation uses a more powerful one.
Enhanced Voice Messaging
WhatsApp voice messages will now arrive as playable voice notes, just as you'd expect. Additionally, WhatsApp will now start only once when Skales opens, preventing unnecessary restarts and ensuring a more stable connection.
Improved Reporting and Feedback
You can now report answers from Iris, the Desktop Buddy, and Lio, not just from the main chat window. This provides a more consistent way to give feedback on AI responses across different parts of Skales.
Smoother Daily Use
We've addressed several daily annoyances. Greetings are now answered without triggering unnecessary tool usage, and the first message on a fresh install receives the same instructions as subsequent messages. Imported skills can be opened reliably, and ChatGPT subscription conversations now correctly reuse cached prompts and report their usage.
New here? Skales is the private AI agent you install like an app - free for personal use, any model, everything stays on your machine.
Download Skales