# Laguna XS 2.1 ![Organization](https://img.shields.io/badge/Org-Poolside-0055ff) ![Total](https://img.shields.io/badge/Total-33B-1f6feb) ![Active](https://img.shields.io/badge/Active-3B-2da44e) ![Context](https://img.shields.io/badge/Context-256K%20tokens-1f6feb) ![Pricing](https://img.shields.io/badge/Price-Free%20%28OpenRouter%29-2da44e) ![License](https://img.shields.io/badge/License-OpenMDW--1.1-d29922) ![Modality](https://img.shields.io/badge/Modality-Text-8957e5) ![Released](https://img.shields.io/badge/Released-Jul%202026-555) > **Quick answer:** Laguna XS 2.1 is Poolside's lightest and fastest agentic coding model — a 33B-total / 3B-active sparse MoE with a 256K context window, trained in-house on 15T tokens. It scores 70.9% on SWE-bench Verified and is small enough to run locally on a Mac with 36 GB RAM. Available free on OpenRouter and Ollama under the commercial-friendly OpenMDW-1.1 license. ## At a Glance **Where Laguna XS 2.1 leads** - Local-first design: 33B total / 3B active — runs on a single DGX Spark or a Mac with 36 GB RAM; Q4_K_M GGUF is only 75 GB. - Strong SWE-bench Verified at 70.9% — competitive with much larger models in its weight class. - Commercial-friendly license: OpenMDW-1.1 allows free commercial and non-commercial use with modification rights. - Native reasoning with preserved thinking between tool calls; per-request enable/disable via `enable_thinking`. - Broad ecosystem support on day one: vLLM, SGLang, TRT-LLM, llama.cpp, Ollama, Transformers, Atomic Chat. - Free on OpenRouter during the beta period. **Where it lags** - Terminal-Bench 2.0 score of 37.5% trails larger agentic models substantially. - Text only — no vision, audio, or multimodal input. - Sliding-window attention (512-token window on 30 of 40 layers) may degrade on very long-range dependencies even within the 256K context. - No user-configurable thinking-effort dial — thinking is binary (on or off), not graduated. **Best for:** On-device and self-hosted agentic coding workflows where low inference cost and hardware portability matter more than absolute frontier performance. ## What Laguna XS 2.1 Is Laguna XS 2.1 is the small tier of Poolside's Laguna model family, released July 2, 2026 on Hugging Face. It is a second-generation architecture, built on lessons from training Laguna M.1, and represents Poolside's first open-weight release. The model was pre-trained in-house on 15 trillion tokens, then post-trained and fine-tuned with reinforcement learning inside Poolside's proprietary `pool` agent harness across 409,000+ agentic and non-agentic training environments. The architecture is a token-choice Mixture-of-Experts transformer with 40 layers: 10 global attention layers and 30 sliding-window attention layers (512-token window), all using sigmoid gating with per-layer rotary scales. There are 256 routed experts plus 1 shared expert per MoE layer, keeping activated compute at ~3B parameters per token despite 33B total weights. The KV cache is quantized to FP8, reducing memory footprint per cached token. A trained speculative decoding draft model (DFlash) is available for lower-latency serving. Laguna XS 2.1 improves over its predecessor Laguna XS.2 by +1.0 point on SWE-bench Verified (70.9% vs. 69.9%) and +5.4 points on SWE-bench Multilingual (63.1% vs. 57.7%). Poolside notes that the gains primarily came from scale, training improvements, and more diverse post-training data — the pre-training corpus is the same as XS.2. The model is available in BF16 on Hugging Face, with FP8, NVFP4, and INT4 quantised variants provided separately. ## Specifications | Field | Value | |---|---| | Organization | Poolside | | Total parameters | 33B | | Active parameters | ~3B per token | | Architecture | Sparse MoE — 256 routed experts + 1 shared, sigmoid gating, per-layer rotary scales | | Layers | 40 (10 global attention, 30 sliding-window attention) | | Sliding window | 512 tokens | | Vocabulary | 100,352 tokens (Laguna family tokenizer) | | Context window | 262,144 tokens (256K) | | Max output | 32,768 tokens | | Numerics | BF16 (FP8/NVFP4/INT4/GGUF variants available) | | Reasoning | Yes — interleaved thinking with preserved thinking | | Input modalities | Text | | Output modalities | Text | | Pre-training data | 15T tokens | | License | OpenMDW-1.1 (commercial use permitted) | | Release date | July 2, 2026 | | HuggingFace | [poolside/Laguna-XS-2.1](https://huggingface.co/poolside/Laguna-XS-2.1) | ## Pricing | | Input (per 1M tokens) | Output (per 1M tokens) | |---|---|---| | OpenRouter (free tier) | $0.00 | $0.00 | | Poolside dedicated API | Contact Poolside | Contact Poolside | Laguna XS 2.1 weights are freely downloadable from Hugging Face under OpenMDW-1.1, which permits commercial use. API access via OpenRouter is free for a limited period during beta; dedicated API pricing for production deployments is available through [platform.poolside.ai](https://platform.poolside.ai/). ## Context Window Laguna XS 2.1 has a **256K-token context window** (262,144 tokens) — roughly **≈ 393 A4 pages** of text in a single request, with a 32K max output. The architecture uses interleaved full global attention (10 layers) and sliding-window attention with a 512-token window (30 layers), which keeps inference efficient while maintaining reasonable long-range coherence. ## Public Benchmark Scores | Benchmark | Score | Source | Date | |---|---|---|---| | [SWE-bench Verified](/benchmarks/swe-bench-verified) | 70.9% | [Poolside HuggingFace model card](https://huggingface.co/poolside/Laguna-XS-2.1) | 2026-07 | | SWE-bench Multilingual | 63.1% | [Poolside HuggingFace model card](https://huggingface.co/poolside/Laguna-XS-2.1) | 2026-07 | | [SWE-bench Pro](/benchmarks/swe-bench-pro) | 47.6% | [Poolside HuggingFace model card](https://huggingface.co/poolside/Laguna-XS-2.1) | 2026-07 | | Terminal-Bench 2.0 | 37.5% | [NVIDIA Build model card](https://build.nvidia.com/poolside/laguna-xs-2.1/modelcard) | 2026-07 | *All Laguna XS 2.1 benchmarking used the Laude Institute's Harbor Framework with Poolside's agent harness, max 500 steps, sandboxed execution, thinking enabled, 256K context. Scores are mean pass@1 averaged over multiple attempts per task. Source: Poolside and NVIDIA published materials. Not Benchgen measurements.* ## Laguna XS 2.1 vs Alternatives | Model | Context | Parameters (total/active) | SWE-bench Verified | SWE-bench Pro | License | |---|---|---|---|---|---| | **Laguna XS 2.1** | 256K | 33B / 3B | 70.9% | 47.6% | OpenMDW-1.1 | | [Laguna S 2.1](/models/poolside/laguna-s-2-1) | 1M | 118B / 8B | — | 59.4% | OpenMDW-1.1 | | [Qwen3-235B-A22B](/models/qwen/qwen3-235b-a22b) | 236K | 235B / 22B | — | — | Apache 2.0 | | [Inkling](/models/thinkingmachines/inkling) | 1M | 975B / 41B | 77.6% | 54.3% | Apache 2.0 | | Qwen3.6-35B-A3B | 35K | 35B / 3B | 73.4% | 49.5% | Apache 2.0 | At 33B/3B, Laguna XS 2.1 is the clear choice for local or constrained-budget deployments where the DGX Spark or equivalent single-node hardware is the constraint — it achieves SWE-bench Verified parity with models 10× its active size, at near-zero inference cost. ## How Laguna XS 2.1 Performs on Real Agent Tasks The 70.9% SWE-bench Verified score is notable for a model with only 3B active parameters: it matches or beats models with 4–10× more active compute on standard coding agent benchmarks. The training methodology — RL inside Poolside's own agent harness across thousands of real environments — produces a model tuned for the "way of working" that matters for unattended agents: verification, persistence, and not declaring victory prematurely. Poolside published full trajectories from its Laguna M.1/XS.2 evaluation at [trajectories.poolside.ai](https://trajectories.poolside.ai/), a transparency standard rare among labs of any size. What the headline numbers don't capture is harness sensitivity. Laguna XS 2.1 is explicitly optimised for Poolside's own `pool` harness; teams running it under a different agent scaffold (different tool schemas, prompt formats, or step limits) should expect some performance degradation relative to published numbers. The Terminal-Bench 2.0 score of 37.5% suggests the model is not yet at frontier level for long-horizon terminal work specifically — the larger Laguna S 2.1 (70.2% on TB 2.1) is the better pick for that workload. ## Use Laguna XS 2.1 via Transformers ```python import torch from transformers import AutoModelForCausalLM, AutoTokenizer model_id = "poolside/Laguna-XS-2.1" tokenizer = AutoTokenizer.from_pretrained(model_id) model = AutoModelForCausalLM.from_pretrained( model_id, dtype=torch.bfloat16, device_map="auto", ) messages = [{"role": "user", "content": "Write a Python retry wrapper with exponential backoff."}] inputs = tokenizer.apply_chat_template( messages, add_generation_prompt=True, return_tensors="pt", enable_thinking=True, # set False to skip chain-of-thought ).to(model.device) outputs = model.generate(inputs, max_new_tokens=1024, do_sample=True, temperature=1.0, top_k=20) print(tokenizer.decode(outputs[0][inputs.shape[-1]:], skip_special_tokens=True)) ``` Or run locally with Ollama in a single command: ```bash ollama run laguna-xs-2.1 ``` ## Related Models & Benchmarks - Compare Poolside family: [Laguna S 2.1](/models/poolside/laguna-s-2-1) - Benchmarks: [SWE-bench Verified](/benchmarks/swe-bench-verified) · [SWE-bench Pro](/benchmarks/swe-bench-pro) · [Terminal Bench](/benchmarks/terminalbench) ## Frequently Asked Questions

What is Laguna XS 2.1?

Laguna XS 2.1 is Poolside's compact agentic coding model: a 33B-total / 3B-active sparse Mixture-of-Experts model with a 256K context window, trained on 15T tokens. It is designed for local deployment and scores 70.9% on SWE-bench Verified.

What is Laguna XS 2.1's context window?

Laguna XS 2.1 has a 262,144-token (256K) context window with a 32K max output — approximately 393 A4 pages of text in a single request.

How much does Laguna XS 2.1 cost?

Laguna XS 2.1 is available free on OpenRouter during Poolside's beta period ($0.00/MTok). The weights are freely downloadable from Hugging Face under the commercial-friendly OpenMDW-1.1 license.

Is Laguna XS 2.1 open source?

The weights are open and downloadable from Hugging Face under the OpenMDW-1.1 license, which permits commercial and non-commercial use with modification rights. It is not OSI open-source (the license includes specific terms), but it is commercially permissive.

Can Laguna XS 2.1 run locally?

Yes. Laguna XS 2.1 is small enough to run on a Mac with 36 GB RAM using a quantized variant (Q4_K_M GGUF is ~75 GB). It is available via Ollama, llama.cpp, and vLLM.

--- *Specs and scores sourced from [Poolside's HuggingFace model card](https://huggingface.co/poolside/Laguna-XS-2.1) and [NVIDIA Build model card](https://build.nvidia.com/poolside/laguna-xs-2.1/modelcard). Last updated 2026-07-23.*

Laguna XS 2.1 — Files