Former OpenAI researchers launch a tool to customize AI models

The startup Thinking Machines Lab has launched its first product, a tool called Tinker designed to simplify the creation of specialized AI models. Will Knight reports for Wired that the company was founded by a prominent group of former OpenAI researchers, including CEO Mira Murati.

Tinker automates the process of fine-tuning powerful open source AI models. This allows developers and researchers to adapt models for specific purposes, such as drafting legal documents or solving math problems. Murati told Wired that the tool aims to make frontier AI capabilities more accessible. The company’s goal is to enable more people to experiment with and conduct research on advanced AI.

Tinker currently supports two open source models: Llama from Meta and Qwen from Alibaba. Users can modify these models using methods like supervised learning, which uses labeled data, or reinforcement learning, which uses feedback. After customization, users can download their new model and use it independently.

The company previously raised $2 billion in seed funding. According to the source article, beta testers praised the tool for being powerful and user-friendly. The company hopes that by making advanced model customization easier, it will encourage more openness in an industry that increasingly relies on closed systems.

Related posts:

Stay up-to-date: