Thinking Machines Releases Inkling 2 Small MoE Model
Thinking Machines Lab has released Inkling 2 Small, a 594B-parameter Mixture of Experts model featuring 17B active parameters.
What happened
According to MarkTechPost, Thinking Machines Lab has officially released Inkling 2 Small. This new machine learning offering is a Mixture of Experts (MoE) architecture boasting a massive 594 billion total parameters. Despite its large overall footprint, it activates only 17 billion parameters during execution, balancing scale with operational efficiency.
The model currently achieves a score of 31 on the Artificial Analysis Intelligence Index. In terms of deployment and accessibility, the release runs on Tinker, and the organization has announced plans to make the model weights available under an Apache 2.0 license.
Why it matters
Mixture of Experts architectures continue to be a focal point for organizations looking to scale model capacity without incurring the massive compute costs of dense models during inference. By activating just 17 billion out of nearly 600 billion parameters, Inkling 2 Small represents this efficiency-focused design philosophy.
Furthermore, the upcoming Apache 2.0 weight release is a notable development for practitioners who prefer open-weights solutions. It provides teams with greater control and flexibility to deploy and modify models locally or within private cloud environments.
What it means for developers and businesses
For developers and small-business owners, the arrival of Inkling 2 Small offers a fresh option to evaluate for production workflows. Running on Tinker and boasting a selective parameter activation design, it targets workloads that demand high capacity paired with optimized active parameter counts.
As the open-weights landscape expands with Apache 2.0 licensing, businesses can increasingly integrate advanced MoE models into their proprietary technical stacks without restrictive licensing hurdles. Developers will want to benchmark its performance on specific domain tasks against existing open models.
What to watch next
Observers should monitor the official release of the Apache 2.0 model weights to see how easily the community can self-host and fine-tune Inkling 2 Small. Additional benchmark evaluations and integration updates with Tinker will also clarify its practical utility for everyday production applications.
Frequently asked questions
What is Inkling 2 Small?
According to MarkTechPost, Inkling 2 Small is a 594 billion parameter Mixture of Experts model developed by Thinking Machines Lab with 17 active billion parameters.
What is the benchmark score for Inkling 2 Small?
The model scores 31 on the Artificial Analysis Intelligence Index, as reported by MarkTechPost.
Will Inkling 2 Small be open source?
Thinking Machines Lab has planned Apache 2.0 open weights for the model, which runs on Tinker.
Sources
This article was written with AI assistance from the sources above. Facts are attributed to their original publishers.
Written by Jayesh Patil
AI & full-stack developer. I build RAG assistants, AI automations and fast websites for businesses.