LANGUAGES Signal 183
Making Local AI Smarter and Faster
↗
Illustration only Photo by Louis Hansel on Unsplash
Kotlin's Junie Local update merges two versions of the Qwen3.6 and Qwen3.8 models to create a single, more capable model that can understand a codebase, do useful work, and finish tasks without long waits.
This update makes local coding agents smarter and faster, enabling developers to use more capable models on their own machines without sacrificing performance. The new model has been tested on multiple public benchmarks and has shown promising results.
Written by elseif from the cluster below · every claim links back to a sourceThe three things worth knowing
The Qwen3.8-3.6-27B-blend model has been shown to complete more tasks than Qwen3.6 while generating 71% fewer output tokens than Qwen3.8.
The model has been tested on multiple public benchmarks, including LiveCodeBench and visual benchmarks, and has shown promising results.
The update also includes experimental NVIDIA support on Windows, allowing developers to run the new model on more machines.
THE READ
What the cluster adds up to.
The Qwen3.8-3.6-27B-blend model is a significant improvement over the original Qwen3.6 and Qwen3.8 models, with the ability to complete more tasks and generate fewer output tokens. This makes it a more practical choice for developers who want to use local AI on their own machines.
The model's performance has been tested on multiple public benchmarks, including LiveCodeBench and visual benchmarks, and has shown promising results. However, further work is needed to improve the model's efficiency and reduce its reliance on reasoning.
The update also includes experimental NVIDIA support on Windows, allowing developers to run the new model on more machines. This expands the reach of the Junie Local update and makes it more accessible to a wider range of developers.
Written by elseif from the cluster below · checked for specifics the sources never containedTHE CLUSTER