Thinking Machines Laboratory, the artificial intelligence company founded by OpenAI exiles has released its first model called Inkling. The up-to-date startup model is open-weight, which means researchers and startups will be able to download and modify it.
In blog entrythe company says Inkling has been trained from scratch to understand audio, video and text input. He says that while Inkling isn’t the best model in popular benchmarks, it performs well on many tasks and is capable of advanced inference and coding. Like many open-weight models, Inkling is relatively immense — 975 billion parameters — and must run on a cluster of specialized chips.
In a sign of how AI models are getting bigger used to create artificial intelligencethe lab also used Inkling technology to refine and refine the software.
This release could facilitate Thinking Machines establish itself as a legitimate player in the hectic and high-spending AI race. Open source models have proven popular because they are cheaper to run than closed source models, which can typically only be accessed for a fee. Open source models can also be more easily modified for different tasks. The best open-weight models currently come from China, but Thinking Machines says the Inkling offers a level of performance similar to those models.
The release of the open-weight model is in line with the vision of artificial intelligence presented by Thinking Machines in: latest blog entry. The company said the technology should not be controlled by a few companies and should be decentralized so more people can build their own models based on their own data.
According to a source familiar with the process who was granted anonymity and the opportunity to speak freely, investigators discovered the strange phenomenon during Inkling’s training. Like other models, it usually provides an explanation of its complicated reasoning in natural language. Inkling decided to forego this in the name of efficiency. “It said the grammar was exaggerated, which is interesting,” the source said. According to this person, the company has brought back natural language reasoning to make model decisions more understandable.
Thinking Machines was founded in February 2025 by several high-profile OpenAI executives and researchers, including Mira Murati, who served as CTO (and briefly CEO) of OpenAI; John Schulman, co-founder of OpenAI, who played a key role in the development of ChatGPT; and Lilian Weng, former vice president of OpenAI, who led the security and robotics efforts.
The startup received the largest seed funding round in history, which initially valued it at $12 billion. Previously, the company released Tinkermodel tuning tool, presented a tool that enables natural voice interactionsand has published research on machine learning.
OpenAI may have started the AI boom with ChatGPT, but defector-led companies like Thinking Machines and Anthropic have entered the space. Anthropic recently filed for an initial public offering that could value the company at more than a trillion dollars. His Claude model has proven popular with many companies, particularly for his coding skills.