NewsBite analysis · AI, Startups
TechCrunch
Thinking Machines wants to build an AI that actually listens while it talks
Thinking Machines Lab, the AI startup founded by former OpenAI CTO Mira Murati, announced on Monday "interaction models" that enable full-duplex conversation—meaning the AI can listen and generate responses simultaneously, similar to a phone call. Current AI models operate in half-duplex mode: the user speaks, the model listens, then responds, then listens again. The company claims its TML-Interaction-Small model achieves a response time of 0.40 seconds, which it says is faster than comparable models from OpenAI and Google. This makes interactions feel more natural and fluid. However, this is a research preview, not a product. A limited research preview will be released in the coming months, with a wider release later this year. The announcement signals a potential shift in how human-AI interactions are designed, but the real-world performance remains to be seen.
MonitorConfidence: MediumShort term horizonExecution risk: High
Published 12 May 2026, 06:40 UTC1 min read
Visit original article ↗Lead image — TechCrunch
Summary
TL;DR
- Thinking Machines Lab announced full-duplex AI interaction models.
- Model TML-Interaction-Small responds in 0.40 seconds.
- Company claims it's faster than comparable models from OpenAI and Google.
- Current AI models are half-duplex; this allows simultaneous processing.
- Founded by former OpenAI CTO Mira Murati.
- Limited research preview coming in months, wider release later this year.
Interesting facts
- The model is named TML-Interaction-Small, suggesting a family of models.
- Full duplex is not new in telephony but novel for AI conversational interfaces.