Laguna XS 2.1 33B MoE Optimized for Agentic Coding
July 3, 2026
Laguna XS 2.1 is a 33B parameter Mixture-of-Experts model specifically tuned for long-horizon tasks. It demonstrates improved performance on SWE-bench Multilingual, targeting agentic coding use cases.
HOW THIS AFFECTS YOU
●
builderYou can deploy this 33B MoE model for coding agents that require long-context reasoning.
●
researcherThe improved SWE-bench Multilingual scores provide a new benchmark for MoE efficiency in coding.