Mirai Murati's Thinking Machines Lab Unveils First AI Model 'Inkling'

Mirai Murati's Thinking Machines Lab has unveiled its first AI model, 'Inkling,' a multimodal model with 975 billion parameters trained on 45 trillion tokens across text, images, audio, and video. The model, released under the Apache 2.0 license on Hugging Face, supports text, image, and audio inputs with a 1 million token context window. Performance benchmarks show Inkling excels in agent tasks, achieving 74.1% on the MCP Atlas compared to NVIDIA's NeMo 3 Ultra, and 77.6% on the SWE-Bench Verify dataset, surpassing NeMo 3 Ultra's 70.7%. However, Chinese models like Z.ai's GLM 5.2 and Kimi K2.6 outperformed Inkling in specific benchmarks. The lab also released a smaller variant, Inkling-Small, with 120 billion active parameters out of 2760 billion total, to be open-sourced after testing. The company, founded by Murati in February 2025 following her departure from OpenAI, raised $2 billion in July 2025 with backing from Andreesen Horowitz and other firms, though a later $5 billion valuation fundraising failed in January 2026.