Welcome Inkling 2 Mini: Advanced Local AI Model Released
The Inkling 2 Mini has been launched as a new local AI model featuring a multimodal Mixture-of-Experts architecture. It boasts approximately 26 billion parameters, with 3 billion active parameters per token, and supports a 1 million token context size. Designed for local setups, it offers significant performance improvements over previous models, particularly in coding and knowledge work tasks.
2 min read1 source
Key takeaways
30-second read- 1Developers can leverage the Inkling 2 Mini for local AI applications, reducing reliance on cloud services and improving response times.
- 2The model's multimodal capabilities allow for both text and image inputs, making it versatile for various use cases in product development.
- 3With its state-of-the-art performance in coding benchmarks, product teams can enhance their tools and applications for better user experiences.
Confirmed
Official source: HuggingFace Blog
Reported
No secondary reporting linked
Verify yourself
Eligibility, local prices and dates for your car, state or utility

The recent launch of the Inkling 2 Mini marks a significant advancement in local AI models, designed specifically for developers and product teams looking for high-performance solutions without relying on cloud infrastructure. This model, developed by Thinking Machines Lab, utilizes a unique Mixture-of-Experts architecture that allows it to function effectively with a smaller physical footprint while maintaining impressive capabilities.
What happened
The Inkling 2 Mini features approximately 26 billion total parameters, with 3 billion active parameters per token. It employs a novel looping mechanism in its architecture, where certain layers are revisited multiple times to enhance processing efficiency. This design allows the model to effectively operate as if it had more layers, achieving a total of 48 logical layers despite having only 24 physical ones. The model supports a context size of up to 1 million tokens and can process both text and image inputs, making it a versatile tool for various applications.
Why it matters
- Local AI Deployment: The Inkling 2 Mini is tailored for local setups, enabling developers to run sophisticated AI tasks without the need for cloud-based solutions, which can lead to faster response times and reduced operational costs.
- Multimodal Capabilities: With the ability to handle both text and images, developers can create more interactive and engaging applications that utilize diverse data types.
- Benchmark Performance: The model has demonstrated superior performance in numerous benchmarks, particularly in coding and knowledge work tasks, which can significantly enhance the productivity of developers and teams working in these areas.
Context and caveats
While the Inkling 2 Mini shows promise, it is essential to note that further validation is needed to fully understand its capabilities compared to other models in the market. The benchmarks indicate strong performance, but practical implementations may vary based on specific use cases and deployment environments. Additionally, the model is shipped in a mixed-precision format, which may require specific hardware configurations to optimize performance.
What to watch next
As developers begin to adopt the Inkling 2 Mini, it will be important to monitor user feedback and performance in real-world applications. The ongoing development of support tools, such as integration with popular frameworks like transformers and MLX, will also be crucial for facilitating its adoption. Furthermore, the potential for future iterations of the model, including those with variable loop configurations, could lead to even greater efficiencies and capabilities.
Buying guide: Best GPU for Running Local LLMs (2026): How to Choose
Sources
1 official- [1]Welcome Inkling 2 Mini, Cutting-edge Performance for your Local SetupOfficial
HuggingFace Blog · huggingface.co · published Nov 19
Newsletter
Weekly AI signal. No spam.


Comments
Sign in to join the discussion
Loading comments…