Daily AI CatchupPoolside's Laguna XS 2.1 Improves Agentic Coding on a Lean 33B Model
Poolside AI released Laguna XS 2.1, a 33B-parameter Mixture-of-Experts model optimized for agentic coding and long-horizon tasks. It improves 5.4 points on SWE-bench Multilingual to 63.1%, ships three quantized checkpoints for resource-efficient deployment, and is available on Hugging Face under the OpenMDW-1.1 license.
Why it matters
💻 Developer · At 33B parameters with quantized checkpoints, this is a realistic self-hosting option if you want agentic coding capability without frontier-model infrastructure requirements.
📦 Product · A capable coding model small enough to self-host changes the build-vs-buy calculus for teams wary of sending proprietary code to a third-party API.
🎨 Design · No direct design impact — this is a developer-tooling model release.
📈 Business · Open, self-hostable coding models at this quality level reduce vendor lock-in risk for engineering-heavy organizations with data residency requirements.
🤔 Just Curious · Poolside released a smaller, open AI coding model that's good enough to run on your own hardware instead of relying on a cloud AI service.
Try this: Pull Laguna XS 2.1 from Hugging Face and benchmark it against your current coding assistant on a real multilingual repo task.
Sources: Poolside: Introducing Laguna XS 2.1