Thinking Machines’ TML-Interaction-Small: Setting New Standards in Real-Time Speech Processing
Thinking Machines presents TML-Interaction-Small with 276B parameters for natural real-time speech interaction. The encoder-free model uses 200ms microturns and demonstrates outstanding cache efficiency. Skepticism grows around TurboQuant while open-source models continue to gain performance rapidly and outpace Moore’s
Insights from China’s AI Labs: Cultural Differences Shape Research
American and Chinese AI labs have similar resources and talent, but differ fundamentally in organizational culture, with American labs driven more by individual career ambition while Chinese teams focus on the overarching goal of optimal model development—a difference with measurable impact on outcomes.
