Poolside AI Introduces Laguna XS.2 and M.1: Agentic Coding Models Reaching 68.2% and 72.5% on SWE-bench Verified

Poolside AI launched Laguna M.1 and Laguna XS.2, two Mixture-of-Experts models built for agentic coding tasks.

Laguna XS.2 is Poolside’s first open-weight model, supports a 131K context window, and can run on local hardware.

The company says the models deliver strong results on coding benchmarks including SWE-bench Verified, SWE-bench Multilingual, and SWE-bench Pro.

Poolside also released its internal terminal-based agent, pool, plus an ACP client-server, as a research preview for long-horizon software work.