Thinking Machines Releases Inkling, Its First Open-Weight Model, Nine Months After Founding

Summary: Thinking Machines Lab, founded by former OpenAI CTO Mira Murati, released its first open-weight model, Inkling, just nine months after the company's founding.

Key facts

• Mixture-of-Experts transformer: 975B total parameters, 41B active, 1M-token context window
• Pretrained on 45 trillion tokens of text, images, audio, and video — natively multimodal
• The company explicitly states Inkling is "not the strongest overall model available today, open or closed"
• Paired with Tinker, Thinking Machines' model-customization platform, enabling organizations to fine-tune Inkling into purpose-built models

Why it matters

Instead of competing head-to-head with GPT-5.6 or Claude Fable 5 on general benchmarks, Thinking Machines is betting that enterprises want a capable, transparent, and customizable foundation — not another locked black box. Building a 975B open-weight model from scratch in nine months is itself a marker of how dramatically AI development timelines have compressed in 2026.

More

• TechCrunch report — TechCrunch
• Official blog — Thinking Machines Lab