Thinking Machines Lab Releases Inkling, 975B Open AI Model
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Thinking Machines Lab debuted Inkling, an open-weight mixture-of-experts AI model with 975 billion total parameters and 41 billion active parameters, designed for broad capability rather than optimization for a single domain.
- Wide outlet coverage including Wired, the Wall Street Journal, Bloomberg, Computerworld, and TechCrunch framed Inkling as Mira Murati's startup's first open-weight model and a major US open-weight release.
- Ghacks reported the model is released under Apache 2.0 licensing, while VentureBeat described Inkling as focused on low cost and resistance to censorship — design choices largely absent from the dominant US-China competitive framing.
- The Decoder, Techstrong.ai, and The Information positioned the launch against Chinese open-source AI dominance, with The Decoder noting Inkling 'leads US labs but trails China,' and Implicator.ai giving it a score of 41.
- Databricks confirmed model availability on its platform, while The Register cast Inkling as a 'frontier AI model that's actually open,' underscoring the release's open-weights accessibility pitch.
Why it matters: Inkling's 975B-parameter scope under Apache 2.0 licensing (per Ghacks) gives US enterprises a domestic, frontier-tier open-weight option, but VentureBeat's noted censorship-resistance design positions it as more than a technical release — a deliberate ideological counterweight to Chinese open-source leaders, per the framing across The Information, Techstrong.ai, and The Decoder.
