Laguna M.1

Laguna M.1 is Poolside’s large sparse Mixture-of-Experts model for agentic coding and long-horizon work. It has 225 billion total parameters, with 23 billion activated per token. Global attention across the layers, a 262,144-token context, and native thinking between tool calls fit its role as a coding agent. The Apache-2.0 license permits commercial and noncommercial adaptation within its terms.

Poolside documents vLLM, SGLang, and Transformers support for M.1, along with FP8 and NVFP4 checkpoints. Its card lists vLLM 0.21.0 or later. Laguna XS.2 is a separate, smaller model aimed at local use.