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

MarkTechPost / 4/29/2026

📰 NewsIndustry & Market MovesModels & Research

Key Points

  • Poolside AI has released two new open-weight agentic coding models, Laguna XS.2 and M.1, designed for long-horizon software engineering tasks.
  • The company reports strong performance on SWE-bench Verified, achieving 68.2% for Laguna XS.2 and 72.5% for M.1.
  • These models target autonomous or agent-style coding workflows, aiming to handle multi-step development requirements beyond short prompts.
  • The release positions Poolside as a competitor in the growing segment of agentic coding LLMs with open-weight availability.
  • The reported SWE-bench improvements suggest practical gains for developers working on complex bug fixing, implementation, and refactoring tasks.

Poolside Releases Laguna XS.2 and M.1: Open-Weight Agentic Coding Models Built for Long-Horizon Tasks

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