The AI startup Thinking Machines Lab, founded by former OpenAI CTO Mira Murati, has released its first open-weight general-purpose AI model, Inkling. The model's architecture draws inspiration from DeepSeek's DeepSeek-V3 and has been optimized during the post-training phase using data from Moonshot AI's Kimi K2.5. Inkling supports deployment on proprietary servers and can be customized and fine-tuned according to specific needs, although its overall performance currently lags behind some leading models. Additionally, Thinking Machines Lab has launched the Tinker platform for enterprises, designed for model fine-tuning and customization. The company is actively expanding its product lineup, having previewed an interactive model capable of processing audio and video inputs in May this year. Founded approximately one year ago, Thinking Machines Lab has completed a seed funding round of $2 billion, achieving a post-money valuation of $12 billion, with investors including Andreessen Horowitz, NVIDIA, AMD, and Jane Street. Notably, the overall usage of Chinese AI models has now surpassed that of their U.S. counterparts.