Thinking Machines releases compact model rivaling flagship AI systems
Thinking Machines introduced Inkling 2 Small, an efficient 17 billion parameter model that performs competitively with much larger systems on web design, coding, and document analysis tasks. The developers plan to open source the model weights after completing safety testing.
The Blend
Artificial intelligence developer Thinking Machines recently introduced Inkling 2 Small, a compact system designed to match larger software models in tasks like coding, visual analysis, and document work. While the entire framework incorporates 594 billion total underlying parameters, the system activates just 17 billion parameters during live operation.
This shift toward efficiency matters because running smaller active models significantly lowers the computing energy and hardware costs needed for complex tasks. According to internal benchmark tests published by Thinking Machines, the new software can write computer code, follow multistep instructions, and generate interactive visual applications like web games entirely from scratch.
The software is currently accessible through a restricted online preview platform. The creators stated that they intend to publish the model files for free public download under an open license after third party safety testing is complete.
While smaller models offer promising cost savings for software creators, it remains uncertain whether light architectures can maintain robust security protections once independent researchers gain full access to the source code.
Written independently by AI News Smoothie from the reporting listed below. Facts belong to the original publishers. Follow the links for their full coverage.
Ingredients
- Introducing Inkling 2 Small - Thinking Machines Lab
Thinking Machines introduced Inkling 2 Small, an efficient model that operates on 17 billion active parameters and will be released under an open license following security checks.