FILE PHOTO: Thinking Machines has unveiled its first product, Tinker, that eases the process of fine-tuning AI models.
| Photo Credit: AP
Thinking Machines, the AI startup founded by former OpenAI CTO Mira Murati, has unveiled its first product, Tinker, that eases the process of fine-tuning AI models. The API has been made available for developers in private beta.
“Tinker brings frontier tools to researchers, offering clean abstractions for writing experiments and training pipelines while handling distributed training complexity. It enables novel research, custom models, and solid baselines,” Ms. Murati said on X while making the announcement.
Normally, the fine-tuning AI models for specific tasks involves managing clusters of GPUs so that training runs are efficient and smooth. Tinker wants to automate this process and give researchers access to the user-friendly API so they can control the different parts of fine-tuning – the loss functions, training loops and data workflows in Python-code, while Tinker takes care of the distributed GPUs.
Tinker has two open-source models, Meta’s Llama and Alibaba’s Qwen for users to fine-tune, a report from Wired stated.
Former OpenAI co-founder Andrej Karpathy commended the release on X saying, “If you’re a researcher/developer, Tinker dramatically simplifies LLM post-training. Compared to the more common and existing paradigm of ‘upload your data, we’ll post-train your LLM,’ this is imo a more clever place to ‘slice up’ the complexity of post-training, both delegating the heavy lifting, but also keeping majority of the data/algorithmic creative control.”
The Lab has been comparatively more transparent than firms like OpenAI while publishing research recently. In September, it posted a blog around the “defeating non-determinism” in large language models (LLMs).
Earlier this year in July, Thinking Machines raised $2 billion in seed funding at a steep valuation of $12 billion.
Published – October 03, 2025 03:16 pm IST
