Poolside opens Laguna S 2.1, a 118B coding model with 1M context

AccBalanced · x · 2026-07-22

Poolside released Laguna S 2.1, its most capable model so far: a 118B total-parameter sparse MoE with 8B active per token, up to 1M context, and thinking/no-thinking modes. The team says it is built for agentic coding and long-horizon work, can run on a single NVIDIA DGX Spark in official NVFP4 quantized form, and is already supported by vLLM out of the box.

Related event: Poolside Releases Laguna S 2.1: 118B Open-Weight Coding Model(28 posts)→

Original post →

More from coding & agent

coding & agent channel →