Laguna
Laguna is Poolside’s agentic coding family. OpenTryOn calls the first-party OpenAI-compatible chat API. The models are text in, text out.
| Registry id | Service | Upstream id | Auth |
|---|---|---|---|
laguna-xs.2 | understand | poolside/laguna-xs.2 | POOLSIDE_API_KEY |
laguna-xs.2.1 | understand | poolside/laguna-xs.2.1 | POOLSIDE_API_KEY |
laguna-m.1 | understand | poolside/laguna-m.1 | POOLSIDE_API_KEY |
XS.2 is 33B / 3B-active (Apache-2.0). XS 2.1 is the successor (OpenMDW-1.1) with native reasoning. M.1 is 225B / 23B-active (Apache-2.0). Pass --laguna-model poolside/laguna-xs-2.1 as an alias for XS 2.1.
Base URL is https://inference.poolside.ai/v1. Override with POOLSIDE_BASE_URL or --base-url.
Thinking is on by default. --no-thinking sends chat_template_kwargs.enable_thinking: false.
opentryon understand --model laguna-m.1 \
--prompt "Write a 3-sentence lookbook caption for a linen trench."
opentryon understand --model laguna-xs.2 \
--prompt "Tighten this PDP copy." \
--system "You edit fashion product copy. Keep claims factual." \
--no-thinking
Images and video are rejected. Use kimi-k2.6, moondream-3.1, or nemotron-omni for media.
Open weights exist, but hosted inference is the Path A integration. There is no local registry id.
MCP tools: understand_laguna_xs_2, understand_laguna_xs_2_1, understand_laguna_m_1. Named-model-only — the understand default stays kimi-k2.6.