Thinking Machines unveils Inkling 2 Small as part of $5 billion fundraising campaign

Thinking Machines Labs Inc. has released a new large language model that can automate artificial intelligence training tasks.

The algorithm, Inkling 2 Small, celebrated his debut late Thursday. Around the same time, Business Insider reported that Thinking Machine is currently raising a $5 billion funding round led by Accel. The company, founded last year by former OpenAI Group PBC chief technology officer Mira Murati, is expected to command a valuation of $40 billion.

Thinking Machines trained Inkling 2 Small on a cluster of GB300 NVL72 racks. Each system contains 72 Blackwell Ultra graphics cards and half as many central processing units. The company completed the initial pre-training phase in 68 hours and then refined the model through a nine-day reinforcement learning workflow. There were also a number of intermediate training steps.

Thinking Machines honed Inkling 2 Small’s programming skills by giving him hundreds of thousands of lengthy programming challenges. The collection of tasks included a series of particularly complicated “autoresearch” exercises. They needed the model to retrain neural networks, optimize LLM components called kernels, and conduct other AI research.

Thinking Machines put Inkling 2 Small’s machine learning capabilities to the test by tasking it with fine-tuning Qwen3.5-4B. The goal was to improve the open source LLM’s ability to make predictions such as weather forecasts. According to the company, Inking 2 Small doubled the model’s prediction value in 4.5 hours.

The LLM is also suitable for automating business tasks. It outperformed Spark Open, Kimi K3, and several other open source models on three popular knowledge work benchmarks. Additionally, higher scores were obtained in several financial task datasets.

Inkling 2 Small includes a vision encoder that can analyze images. In addition, it allows the LLM to visually check the quality of the graphics it generates. When a user asks the model to create an eCommerce store, it can not only check the website’s code for errors, but also visually inspect product pages to verify that they are displayed correctly.

The vision encoder accounts for 292 million of Inking 2 Small’s 594 billion parameters. About 17 billion of these parameters are activated when a user enters a prompt. Inking 2 Small can adjust the processing power spent on tasks, allowing developers to trade output quality for greater cost efficiency, or vice versa.

Advanced users can extend the model’s capabilities by training it afterward or by customizing it using their own data. According to Thinking Machines, the Inkling 2 Small learns faster than its previous models. This should lead to more cost-effective post-training runs.

Inkling 2 Small will be available at launch via Thinking Machines’ Tinker cloud service. The company plans to release the model’s weights under an open source license next week.

Image: Unsplash