Thinking Machines Lab ships its first model and argues interactivity is what OpenAI gets wrong about voice

Editor's take

Thinking Machines Lab has launched its inaugural AI model, a multimodal system designed for near real-time interaction that diverges from traditional conversational paradigms.

This development is significant as it directly challenges the prevailing question-and-answer framework in voice AI, a space dominated by incumbents like OpenAI's GPT Realtime 2 and Google's Gemini. By processing audio, video, and text concurrently in 200-millisecond segments, Thinking Machines aims to enable a more fluid and natural human-computer interface, potentially unlocking new applications in areas requiring immediate contextual understanding.

Future progress will hinge on the model's actual latency and the breadth of its contextual comprehension across modalities. Demonstrating sustained, natural interaction that surpasses current benchmarks in responsiveness and intuitive understanding will be key to validating its proposed advantage over existing solutions.