Thinking Machines Releases TML-Interaction-Small 276B-A12B: Real-Time Voice AI That Kills Standard VAD

Summary

Thinking Machines released TML-Interaction-Small 276B-A12B alongside a research paper titled "Interaction Models: A Scalable Approach to Human-AI Collaboration." The model is a mixture-of-experts architecture with 276 billion total parameters and 12 billion active parameters, designed specifically for real-time human-AI interaction rather than traditional turn-based conversation. The system achieves sub-200ms latency per response cycle and eliminates the need for standard Voice Activity Detectio