Poolside releases Laguna S 2.1, the open-weight coding model pitched as the West’s answer to DeepSeek and Qwen

San Francisco AI lab Poolside has released Laguna S 2.1, an open-weight coding model designed for agentic tasks. This 118-billion-parameter Mixture-of-Experts system activates 8 billion parameters per token and supports a 1 million token context window. According to company benchmarks, Laguna S 2.1 matches or exceeds the performance of rivals several times its size. The model is compact enough for a single Nvidia DGX Spark desktop system, with weights available on Hugging Face.