T
traeai
Sign in
返回首页
lmarena.ai(@lmarena_ai)

Laguna XS.2 & M.1 by @poolsideai are ready in the Code Arena: Front-end. Come bring your toughe...

5.2Score
Laguna XS.2 & M.1 by @poolsideai are ready in the Code Arena: Front-end.

Come bring your toughe...

TL;DR · AI Summary

Poolside AI 发布开源 MoE 编程模型 Laguna XS.2(33B/3B),专为智能体式前端开发任务优化,支持单卡运行,Apache 2.0 协议。

Key Takeaways

  • Laguna XS.2 是 Poolside 自研的 33B 总参、3B 激活的稀疏 MoE 编程模型
  • 首次开源、支持单 GPU 部署,聚焦 agentic webdev 长周期任务
  • 已上线 Arena.ai Code Arena 前端赛道,开放用户评测与投票

Outline

Jump quickly between sections.

  1. Arena.ai 宣布 Laguna XS.2 与 M.1 模型上线 Code Arena 前端赛道。

  2. Laguna XS.2 定义为面向 agentic webdev 的长周期编程任务专用模型。

  3. 33B 总参数、3B 激活 MoE 架构,全栈自研,单卡可运行。

  4. Apache 2.0 协议开源,权重已公开,强调可商用与可部署性。

  5. 通过 Arena.ai 投票制评估输出质量,正式分数尚未公布。

Mindmap

See how the topics connect at a glance.

查看大纲文本(无障碍 / 无 JS 友好)
  • Laguna XS.2 开源编程模型
    • 架构特性
      • 33B 总参 / 3B 激活 MoE
      • 单 GPU 可运行
    • 应用场景
      • Agentic webdev
      • 长周期代码生成
    • 发布信息
      • Apache 2.0 开源
      • Code Arena 前端赛道上线

Highlights

Key sentences worth saving and sharing.

  • Laguna XS.2 是 Poolside 的第一款 open-weight 模型,33B total / 3B active MoE,专为 agentic coding 和长视野任务设计。

    Quote from @poolsideai

    ⬇︎ 下载 PNG𝕏 分享到 X
  • 训练完全在 Poolside 自建技术栈上完成,不依赖第三方基础设施或数据集。

    Quote from @poolsideai

    ⬇︎ 下载 PNG𝕏 分享到 X
  • Released under Apache 2.0 — 可自由使用、修改、分发,包括商业用途。

    Quote from @poolsideai

    ⬇︎ 下载 PNG𝕏 分享到 X
#Poolside AI#MoE#代码生成#开源模型#前端开发
Open original article

Come bring your toughest agentic webdev tasks and vote for the outputs that deliver best for your use case.

Scores coming soon. https://t.co/0yx9lGoapX" / X

Image 1: Square profile picture

Arena.ai

@arena

Laguna XS.2 & M.1 by

@poolsideai

are ready in the Code Arena: Front-end. Come bring your toughest agentic webdev tasks and vote for the outputs that deliver best for your use case. Scores coming soon.

Image 2: Image

Quote

poolside

@poolsideai

Apr 28

Today we’re releasing Laguna XS.2, Poolside’s first open-weight model. It’s a 33B total / 3B active MoE model built for agentic coding and long-horizon tasks. Trained fully in-house on our own stack. Runs on a single GPU. Released under Apache 2.0. Links Image 3: 👇 Weights:

Image 4: Image

8:41 PM · May 1, 2026

7,732 Views

AI may generate inaccurate information. Please verify important content.