Wang Li's Startup's Large Model Debut, Saying Goodbye to a "0 Model with a $12 Billion Valuation"

▤ Summary

Thinking Machines Lab (TML), founded by former OpenAI leaders, released its first interactive model, TML-Interaction-Small, aimed at real-time human–AI interaction. The model targets ~200ms response latency and claims interaction latency about four times faster than GPT-realtime-2.0, though its raw model intelligence still trails top GPT modes in some tests. The team highlights deep engineering and compute partnerships and says larger-scale models and deployments are planned, but admits engineering pressure and model completeness remain challenges. The company has also drawn attention for rapid fundraising and a high valuation, prompting questions about product maturity versus market expectations.

▦ Linked Companies

NVIDIA Corporation

$1B+

Thinking Machines Lab

$1M to $5M