Respan: Span-01 Lite (free)

Respan: Span-01 Lite (free) is a behavior scoring model from Respan, released on 26 September 2026. It sits in the decisions category and is the free, lighter tier of Span-01. Rather than generating conversational text, it is built to return a probability for each behavior you describe in plain language, which makes it a scoring component you place inside a larger system rather than an assistant you chat with.

The model takes text as its input and produces decisions as its output. You supply plain-language definitions of the behaviors you care about, and for each one the model returns the probability that the behavior is present. Its context length is not published, and no request parameters are listed as supported. The maximum output is recorded as zero, which reflects that this endpoint returns scored decisions rather than generated prose. There is no image, audio or other non-text input, and no text generation on the output side.

Because the output is a probability per defined behavior, the practical uses follow from scoring rather than writing. You can use it to triage or route incoming text, to flag content that matches a behavior you have defined, to run lightweight moderation or compliance checks over user submissions, or to enrich a dataset with per-behavior labels. It also fits as a cheap first pass in a pipeline where only the borderline cases get escalated to a heavier model. Anywhere you would otherwise write a classifier or hand-label examples, this model lets you express the target behavior in words and get a probability back.

You call it through api14's OpenAI-compatible API using the model id "respan/span-01-lite:free". Because the interface follows the OpenAI chat completions shape, you can point an existing OpenAI client at api14 and change only the model id. The model accepts text input and returns decisions; no additional request parameters are listed as supported, so keep your requests to the basics and read the returned probabilities rather than expecting free-form text. The "free" suffix in the model id indicates the free tier.

The main thing to plan around is that this is the Lite tier. If your behavior definitions are subtle, or you need scoring quality beyond what the lighter tier provides, the full Span-01 model from the same provider is the natural next step. Since the context length is not published and no parameters are listed, treat this as a focused scoring endpoint and validate it against your own examples before depending on it in production.