The Thinking Machines Lab, led by former OpenAI CTO Mira Murati, has released its first open AI model, Inkling.
Today, we are introducing Inkling.
Inkling reasons efficiently across text, image, and audio modalities. We are making the full weights available.https://t.co/Ghebq5mG30
Available today for fine-tuning on Tinker. Play with it in the Inkling Playground. 🧵— Thinking Machines (@thinkymachines) July 15, 2026
The neural network is built on the MoE architecture: only part of the network is activated for each query, speeding up inference without losing depth. It has a total of 975 billion parameters, with 41 billion active. The context window is 1 million tokens. The model was trained from scratch on 45 trillion tokens of text, images, audio, and video.
Inkling can reason across multiple modalities but currently only provides text-based responses, including code and structured data. Users can adjust the balance between speed and quality using a controllable reasoning level.
Alongside Inkling, developers introduced a lighter version, Inkling-Small, with 276 billion parameters, 12 billion of which are active. The full weights of this model have not yet been released, as the company continues testing.
Where Inkling Excels
The model has shown impressive results in agent tasks. On MCP Atlas, Inkling scored 74.1%, nearly 30 points higher than Nvidia Nemotron 3 Ultra, its main Western competitor among open models. On SWE-Bench Verified, it scored 77.6% compared to Nemotron’s 70.7%.
On FORTRESS Adversarial, Murati’s project scored 78%. It is currently the most powerful open model released by a Western company, though it still trails behind Chinese competitors.
On Terminal Bench 2.1, Inkling scored 63.8%. Z.ai GLM 5.2 scored 82.7%. The recently released Kimi K3 achieved 88.3%. It also leads in Humanity’s Last Exam.
Thinking Machines positions Inkling as a “universal” model: no sacrifices in some tasks for superiority in others. The focus is on customization, with the company believing businesses value a flexible model tailored to specific tasks.
Previously, Murati left OpenAI in September 2024, citing a desire to pursue her own research.
