Poolside releases Laguna-S-2.1, a new 120B parameter open-weights model
Poolside has released Laguna-S-2.1, a new open-weights large language model featuring 120 billion parameters. The release includes GGUF versions that are currently compatible with a custom fork of the llama.cpp inference library. This release introduces a significant new contender in the high-parameter open-weights category, which is highly anticipated by the local LLM community. It offers developers and researchers a powerful new option for running large-scale models locally. To run the GGUF versions of Laguna-S-2.1, users must currently use a custom fork of llama.cpp rather than the main repository. The model weights and GGUF files are hosted on Hugging Face.
## BACKGROUND
Poolside AI is a startup focused on building large language models optimized for software engineering and coding tasks. GGUF is a binary file format designed by the llama.cpp project for fast saving and loading of model data, making it highly efficient for local LLM inference. llama.cpp is a popular open-source C/C++ library that enables LLM execution on consumer hardware.