VercelEngineering PostOfficial Source
Laguna S 2.1 is now available on AI Gateway
Poolside’s open-weight Laguna S 2.1 is now callable through Vercel AI Gateway in free 256K and paid 1M-context variants, with thinking and no-thinking modes for coding agents.
Vercel · Jul 21, 2026
Source Summary
**Laguna S 2.1** is an open-weight Mixture-of-Experts model available through AI Gateway. The free endpoint has a **256K context window**; the paid endpoint supports **1M tokens** and the model offers thinking and no-thinking modes.
Practical Implication
Builders can test the free model ID before moving long-running coding, browser-tooling, MLOps, or research tasks to the paid endpoint. In thinking mode, Vercel reports scores of **70.2%, 78.5%, and 59.4%** on Terminal-Bench 2.1, SWE-bench Multilingual, and SWE-Bench Pro respectively.
Agent-Ready Context
**Laguna S 2.1** is an open-weight Mixture-of-Experts model available through AI Gateway. The free endpoint has a **256K context window**; the paid endpoint supports **1M tokens** and the model offers thinking and no-thinking modes. Builders can test the free model ID before moving long-running coding, browser-tooling, MLOps, or research tasks to the paid endpoint. In thinking mode, Vercel reports scores of **70.2%, 78.5%, and 59.4%** on Terminal-Bench 2.1, SWE-bench Multilingual, and SWE-Bench Pro respectively. Those benchmark figures come from the release material and do not establish latency, cost, or performance on a specific repository. The free and paid variants also differ substantially in context capacity.
Context Map
modelcodingresearch#open-models#coding-agents#model-selectionUncertainty
Those benchmark figures come from the release material and do not establish latency, cost, or performance on a specific repository. The free and paid variants also differ substantially in context capacity.