Thinking Machines bets on efficiency over size with its second model, Inkling Small
What it does
Thinking Machines, the AI lab founded by Mira Murati, former OpenAI CTO, has launched Inkling Small. This new model is a leaner version of their original Inkling, with less than a third of its size. Despite being smaller, Inkling Small delivers stronger performance on several coding and reasoning benchmarks. It is open-weights, meaning developers get transparent access to the model’s internal parameters without proprietary restrictions.
Why it matters
Inkling Small signals a shift away from the era of simply scaling AI models bigger and bigger. Instead, Thinking Machines focuses on efficiency and smarter architecture to maximize accuracy and reasoning power while shrinking computational load. This creates opportunities for real-world applications where speed, cost, and resource constraints matter more than raw model size. It challenges the assumption that bigger models are always superior and highlights how targeted improvements in model design can tighten performance per parameter.
Who it is for
Developers and AI builders who need powerful reasoning models but lack the infrastructure to run large, resource-heavy models will find Inkling Small appealing. It suits coding assistants, automated reasoning tools, and businesses seeking to embed compact yet capable AI without huge cloud spend. Researchers interested in open-weight models can probe and adapt Inkling Small to fit specialized needs.
The catch
While smaller and more efficient, Inkling Small might not outright replace larger foundation models in every scenario. Some complex and large-scale tasks may still benefit from the scope and breadth of bigger models. It also remains to be seen how it performs across diverse real-world benchmarks beyond coding and reasoning, and how it integrates with popular AI platforms and toolchains.
What to watch next
Adoption rates by developers and startups will reveal how much demand exists for smaller, open-weights models like Inkling Small. Watch for improvements or spin-offs that push efficiency even further or broaden capabilities beyond coding assist and logic tasks. Also monitor Thinking Machines’ moves on partnerships and ecosystem support, which will be critical to wider impact and commercial traction.
AI Quick Briefs Editorial Desk
