Poolside: Laguna XS 2.1 (free)

Poolside: Laguna XS 2.1 (free) is a text model from Poolside, listed on api14 under the model id poolside/laguna-xs-2.1:free. It is the latest coding agent model in the 33B-A3B category and a step forward from Poolside's Laguna XS.2 model, which was released in April 2026. The provider release date for Laguna XS 2.1 is 2026-07-02. If you are choosing a model for coding agent work and want a Poolside option in this size class, this is the current entry in that line.

The model is text in, text out. It accepts text input and produces text output, and it does not take images, audio or other modalities. The context window is 262144 tokens, which is large enough to hold substantial source files, long conversation histories or extended agent transcripts in a single request. The maximum output is 32768 tokens, so a single response can be long, but it is bounded well below the context window. Supported request parameters are include_reasoning, max_tokens, reasoning, temperature, tool_choice and tools.

Those capabilities point to a fairly specific set of uses. The large context window suits repository-scale tasks where you need to pass many files or a long history of edits and observations. The tools and tool_choice parameters make it a natural fit for agent loops that call functions, run commands or query external services, and the reasoning and include_reasoning parameters let you control and inspect the model's reasoning output. The 32768 token output limit supports multi-file diffs, long explanations or structured plans in one response. Typical work includes code generation and refactoring, code review, debugging with tool access, and agent workflows that need to carry a lot of context between steps.

You call it through api14's OpenAI-compatible API using the model id poolside/laguna-xs-2.1:free. Because the endpoint follows the OpenAI chat completions shape, you send a request with a messages array and set model to that id. Alongside model and messages, you can pass max_tokens to cap the response length, temperature to adjust sampling, tools to declare callable functions and tool_choice to steer whether and which tool is used. The reasoning and include_reasoning parameters are also accepted if you want to enable reasoning behaviour or have reasoning content returned. Note that the model id carries a :free suffix, so it is the free variant listed on api14.

Keep the input and output modalities in mind when deciding. If your task depends on image, audio or other non-text input, this model will not handle it, and a multimodal model would fit better. The 32768 token output ceiling also means very large single responses may need to be split across calls or handled in stages. Within those bounds, Laguna XS 2.1 is aimed at coding and agent workloads where text context and tool use are the main requirements.