Thinking Machines Lab, led by former OpenAI CTO Mira Murati, has launched a new AI model that eliminates pause-driven interactions by enabling simultaneous listening, seeing, and speaking, mimicking natural human conversation.
- New dual-model architecture supports real-time interaction with sub-0.4 second latency.
- Multimodal system processes audio and visual signals without heavy encoders.
- Designed to enhance high-stakes and enterprise AI applications with fluid conversation.
What happened
Thinking Machines Lab, founded by former OpenAI CTO Mira Murati, has introduced a novel class of AI interaction models aimed at transcending traditional turn-based AI communication. Unlike conventional systems which require users to complete their inputs before AI processes and responds, this new architecture enables simultaneous input and output processing.
The company’s innovative approach involves breaking down communication into micro-turns of about 200 milliseconds, creating a seamless flow of interaction. The architecture pairs a fast interaction model that handles dialogue and immediate responses with an asynchronous background model responsible for complex reasoning and external data fetching to support the conversation dynamically.
Why it matters
Current AI models force users into unnatural conversational patterns due to their inability to process interruptions or subtle conversational cues in real time. Thinking Machines’ new model promises to overcome these limitations by enabling truly humanlike back-and-forth communication, which is vital for AI to serve as a collaborative partner in sensitive or complex situations such as medical surgery or industrial monitoring.
The significant reduction in latency, demonstrated by sub-0.4 second turn-taking times outperforming competitors like Google’s Gemini and GPT-based models, suggests a leap forward in AI responsiveness. This improved interaction fidelity will enhance user experience in customer service, operational safety monitoring, and other enterprise settings that demand real-time awareness and reaction.
What to watch next
Future developments will likely focus on expanding the capabilities of these interaction models in real-world environments, especially integrating them into applications requiring multimodal perception and instant feedback. Observers should monitor how Thinking Machines’ technology scales, particularly in demanding fields like healthcare assistance and manufacturing safety.
Additionally, the effectiveness of this dual-model system to blend deep reasoning and live interaction will be crucial. Its adoption and performance in commercial AI platforms could set new standards for conversational AI speed, naturalness, and utility, positioning Thinking Machines as a potential leader in next-generation AI interfaces.