Baseten: Elephant 1 (free)

Elephant 1 is a decision model from Baseten, released under the model id baseten/elephant-1:free. It comes from Base Labs, Baseten's research team, and is built on Qwen3.5-9B, a 9B parameter base. The model's main category is decisions, which reflects what it is designed to do rather than general chat or content generation. The provider release date listed for this model is 2026-10-21.

The model takes text as its input modality and produces decisions as its output modality. It is built to work from a state, given as text or JSON, together with typed questions. Those questions can be yes/no, multiple choice, or ordinal. The context window is 16,384 tokens, and the maximum output is 14,745 tokens. Supported parameters are not listed for this model, so you should not assume that sampling controls such as temperature or top_p are available when you call it.

Because the output is a decision rather than free-form prose, the practical uses follow from asking structured questions about a supplied state. You can pass a JSON state and ask a yes/no question to get a binary judgment, or present a set of options and ask the model to choose among them. Ordinal questions let you place an outcome on a scale rather than pick a single label. This suits workflows such as routing or triage, where a request or record needs to be classified before further processing, and screening steps where a state must be checked against a condition. It also fits any pipeline where you want a compact, typed answer that downstream code can act on, instead of text that has to be parsed.

You call the model through api14's OpenAI-compatible API using the model id "baseten/elephant-1:free". Because the API is OpenAI-compatible, you send requests in the usual chat completions shape, with the state and your typed questions in the message content, and you read the decision from the response. The model accepts text input and returns decisions, so keep the prompt focused on the state and the question you want answered. Since no supported parameters are listed, send the model id and your messages and avoid relying on extra request parameters to change behaviour.

Keep the context limit in mind: 16,384 tokens covers the state plus your questions, so very large documents will need to be trimmed or split before you send them. The output cap of 14,745 tokens is generous for a decision, but the answer itself is a decision rather than an explanation, so if you need long-form reasoning, summarisation, or open-ended generation, a general text model may fit better. Elephant 1 is the right choice when the task is to answer typed questions about a given state and you want the result as a decision.