- Full-duplex voice interaction enables natural conversations with AI agents
- Qoder Voice debuts globally as AI coding shifts beyond text interfaces
Alibaba-backed AI coding platform Qoder has launched Qoder Voice, describing it as China’s first real-time voice AI agent that allows users to converse naturally with an AI assistant while it executes tasks in the background.
Integrated into Qoder’s Quest mode, the feature lets users summon a persistent voice assistant from any interface.
Instead of issuing one-off voice commands, users can engage in continuous, two-way conversations with the agent, refining requests, interrupting responses and adjusting tasks as they would with a human assistant.
Unlike traditional push-to-talk voice interfaces, which simply convert speech into text commands, Qoder Voice supports full-duplex interaction.
Once a request is made, the system interprets the user’s intent, creates a task, invokes the necessary tools and executes them in parallel, while allowing the conversation to continue uninterrupted.
For example, when asked to refactor code, the agent can propose multiple approaches. If the user responds, “Option B looks better, but add caching,” the agent immediately revises its plan and continues execution without restarting the workflow.
Qoder Voice has launched first on the platform’s international edition for individual and enterprise subscribers, with a three-day free trial. A China version is expected to follow soon.
Voice collaboration
The launch reflects a broader shift in AI productivity tools from text-based prompting toward more natural, conversational workflows.
As AI agents take on increasingly complex, multi-step tasks, voice is emerging as a more efficient interface, particularly when users are mobile or otherwise unable to type.
Real-time, full-duplex conversation also enables AI agents to participate in problem-solving rather than simply responding to sequential commands.
China’s AI agent market reached 80.4 billion yuan ($11.87 billion) in 2025, up 123.2% year on year, according to iiMedia Research.
AI coding platforms are increasingly evolving from text-centric integrated development environments into collaborative agent platforms.
Why it matters globally
Qoder Voice is powered by Qwen-Audio-3.0-Realtime, which Alibaba says ranks first globally on Artificial Analysis’ Speech-to-Speech Index with a score of 84.1%.
The launch underscores how competition in AI coding assistants is expanding beyond model performance to user interaction.
As voice interfaces become capable of sustaining contextual conversations while executing tasks, they could lower the barrier to software development, allowing users to build applications through natural speech rather than traditional coding workflows.
If widely adopted, that shift could reshape how developers — and eventually non-programmers — interact with software creation tools.
