Mira Murati releases open-source model Inkling

Mira Murati released Inkling on July 15, a 975-billion-parameter multimodal model with full weights on Hugging Face under an Apache 2.0 license and fine-tuning on Tinker.

Thinking Machines Lab released Inkling on July 15 and published the full model weights on Hugging Face under an Apache 2.0 license. The company also offers fine-tuning on its Tinker cloud platform.

Inkling is a multimodal model trained from scratch. It has 975 billion total parameters and uses a mixture-of-experts architecture that activates about 41 billion parameters per task. The model accepts text, images, audio and video. Its context window is around one million tokens, roughly 750,000 words. Thinking Machines says Inkling was pretrained on about 45 trillion tokens across multiple data types.

The company reported Inkling’s scores on several agentic benchmarks. On MCP Atlas Inkling scored 74.1 percent. On SWE-Bench Verified it scored 77.6 percent. On FORTRESS Adversarial it scored 78.0 percent. In the company’s comparisons, Inkling outscored Nvidia’s Nemotron 3 Ultra by about 30 points on MCP Atlas and by about 7 points on SWE-Bench Verified.

Other models scored higher on some tests. Z.ai’s GLM 5.2 posted 82.7 percent on Terminal Bench 2.1 while Inkling registered 63.8 percent on the same test. Kimi K2.6 led on Humanity’s Last Exam. Thinking Machines acknowledged Inkling is not the top performer on every metric and described the model as an open-weights option developed outside China.

Thinking Machines previewed Inkling-Small, a 276-billion-parameter variant with about 12 billion active parameters. The company said the smaller variant matches the larger model on most reasoning benchmarks and that its weights will be released after internal testing, with no timeline provided.

Mira Murati left OpenAI in September 2024 and founded Thinking Machines Lab in February 2025. The company raised $2 billion in July 2025 at a reported $12 billion valuation and later pursued a larger fundraising effort that did not close by January 2026.

Murati wrote on X: “Our first model, Inkling. Trained from scratch, weights are open, fine-tunable on Tinker today.”

The material on GNcrypto is intended solely for informational use and must not be regarded as financial advice. We make every effort to keep the content accurate and current, but we cannot warrant its precision, completeness, or reliability. GNcrypto does not take responsibility for any mistakes, omissions, or financial losses resulting from reliance on this information. Any actions you take based on this content are done at your own risk. Always conduct independent research and seek guidance from a qualified specialist. For further details, please review our Terms, Privacy Policy and Disclaimers.

Articles by this author