README.md

---
title: README
emoji: 📈
colorFrom: blue
colorTo: indigo
sdk: static
pinned: false
---

<p align="center">
  <img alt="poolside-banner" src="https://poolside.ai/assets/laguna/poolside-banner.svg" width="800px">
</p>

Welcome to the official Hugging Face organization for Poolside’s open models.

*Visit our [models page](https://poolside.ai/models) to learn more about model building at Poolside.*

> [!NOTE]
> **New**: read our [technical report](https://poolside.ai/assets/laguna/laguna-m1-xs2-technical-report.pdf) on Laguna M.1 and Laguna XS.2.

---

## Laguna S 2.1

[Our most recent release](https://huggingface.co/poolside/Laguna-S-2.1): Laguna S 2.1 (118B-A8B), designed for agentic coding and long-horizon work. Laguna S 2.1 outperforms larger coding agent models, scoring 40.4% on DeepSWE. Free to use under OpenMDW-1.1.

*[Release blog post](http://poolside.ai/blog/introducing-laguna-s-2-1)*.

---

## Laguna XS 2.1

[Our efficient small model](https://huggingface.co/poolside/Laguna-XS-2.1): Laguna XS 2.1 (33B-A3B), designed for agentic coding and long-horizon work on a local machine. It uses Sliding Window Attention with per-head gating in 30 out of 40 layers for fast inference and low KV cache requirements.

*[Release blog post](http://poolside.ai/blog/introducing-laguna-xs-2-1)*.

---

## Laguna M.1

[Our first flagship model, now available open weight](https://huggingface.co/poolside/Laguna-M.1): Laguna M.1 (225B-A23B) is available in base and post-trained (BF16, FP8 and NVFP4) variants under Apache 2.0 licenses. 49.2% on SWE-Bench Pro.

*[Release blog post](https://poolside.ai/blog/laguna-a-deeper-dive)*.

---