OpenAI Updates ChatGPT Desktop App with Voice-Controlled Multi-Step Task Execution
OpenAI has updated its ChatGPT desktop application to support ChatGPT Voice powered by the GPT-Live model series, enabling users to control AI agents and execute complex, multi-step tasks on their computers. This update allows the desktop assistant to perform actions like creating code threads, submitting pull requests, and accessing screen content on macOS. This release represents a significant shift from simple voice conversation to active, OS-level task execution, bringing AI agents closer to daily workflows. It also intensifies competition with Anthropic, which recently updated Claude's voice mode to perform tasks across various productivity applications. The desktop voice feature integrates with ChatGPT Work and Codex, and on macOS, it utilizes Appshots to read on-screen content including alternative text. In contrast to the mobile version which focused on fluid conversation, the desktop version is designed for complex execution and can prompt users for information or confirmation when needed.
## BACKGROUND
The update relies on GPT-Live, OpenAI's advanced voice model designed for natural human-AI interaction, and Codex, an AI system tailored for automating software engineering tasks. Meanwhile, Anthropic competes in this space using its Claude model family, which includes the high-capability Opus, balanced Sonnet, and lightweight Haiku models.