← Back

Wire · technology

Poolside drops Laguna S 2.1, an open-weight coding model that beats rivals 10x its size

Published

22 July 2026

Topic

technology

Sectors

AI Frontier ModelsAI Infrastructure

Geography

United States

Source

Read at venturebeat.com

Verified

Fusion42 · 22 July 2026 · Fusion42 review

Poolside released Laguna S 2.1, a 118B-parameter sparse coding model that beats rivals 10x its size on benchmarks while running on minimal inference compute. The company shipped it in under nine weeks as an open-weight alternative to Chinese models, positioning sparse architecture as the cost-efficient path for enterprise AI agents.

This Wire brief sits within Fusion42's coverage of AI Frontier Models and AI Infrastructure.

◆ The Wire takeaway

If you're building enterprise AI agents or deploying code models behind security walls, Poolside just proved that sparse architecture cuts inference costs enough to run on a desktop. Your customers who rejected closed APIs because they need to self-host now have a Western alternative that matches larger closed models — and that changes which supplier gets called first.

Coverage

1 source · 22 Jul 2026

Related on Wire

Topics

AI Frontier ModelsAI Infrastructureopen-weight-modelscoding-agentssparse-architecturemoe-inferenceenterprise-aigeopolitical-competition