Everyone an Edison: Ex-OpenAI CTO launches Tinker to democratize AI model fine-tuning

Thinking Machines Lab, the heavily funded AI startup cofounded by former OpenAI CTO Mira Murati, has launched Tinker, a tool that automates the creation of custom frontier AI models through fine-tuning. The product represents the company’s bet that democratizing access to advanced model customization will be the next major frontier in artificial intelligence development.

What you should know: Tinker allows businesses, researchers, and hobbyists to fine-tune cutting-edge AI models without managing complex infrastructure or specialized software tools.

• Users can currently fine-tune two open source models: Meta’s Llama and Alibaba’s Qwen through supervised learning or reinforcement learning methods.
• The tool abstracts away distributed training complexities while giving users full control over data and algorithms.
• Applications open Wednesday, with free API access initially before eventual monetization.

In plain English: Fine-tuning is like teaching an AI model new skills by giving it specialized training data—similar to how a general doctor might take additional courses to become a heart surgeon. Traditionally, this process required expensive computer clusters and complex software that only big companies could afford. Tinker makes this accessible to anyone by handling the technical complexity behind the scenes.

The powerhouse team: Thinking Machines Lab assembled a roster of prominent OpenAI veterans who played core roles in creating ChatGPT.

• Mira Murati previously served as OpenAI’s CTO and briefly as CEO during Sam Altman’s temporary ouster in late 2023.
• Cofounder John Schulman led the reinforcement learning work that fine-tuned ChatGPT’s underlying language model.
• Other cofounders include Barret Zoph (ex-VP of research), Lilian Weng (safety and robotics), Andrew Tulloch (pretraining), and Luke Metz (post-training).

The massive funding: The startup raised $2 billion in seed funding in July, achieving a $12 billion valuation before launching any products.

Why this matters: Current fine-tuning requires acquiring GPU clusters and managing complex software tools, limiting access to major companies and academic institutions.

• “We’re making what is otherwise a frontier capability accessible to all, and that is completely game changing,” Murati says. “There are a ton of smart people out there, and we need as many smart people as possible to do frontier AI research.”

Early user feedback: Beta testers highlight Tinker’s combination of simplicity and powerful capabilities compared to existing alternatives.

• Eric Gan from Redwood Research, a company focused on AI risks, uses Tinker’s reinforcement learning to tune models for specialized security research, noting it’s “definitely much simpler than doing the RL from scratch.”
• Robert Nishihara, CEO of Anyscale, a company that supplies technology for managing large-scale AI projects, says while other tools like VERL and SkyRL exist, “Tinker offers a remarkable mix of abstraction and tunability.”

The bigger picture: Tinker’s launch comes as most US AI companies increasingly keep their best models closed, while China leads in open source frontier AI development.

• Murati hopes the tool will reverse the trend toward closed commercial AI models and bridge the growing gap between frontier labs and academic researchers.
• The company currently vets API access but plans automated systems to guard against misuse of the fine-tuning capabilities.

What they’re saying: The team emphasizes their commitment to democratizing advanced AI capabilities.

• “There’s a bunch of secret magic, but we give people full control over the training loop,” Schulman explains. “We abstract away the distributed training details, but we still give people full control over the data and the algorithms.”