Respan: Span-01

Respan: Span-01 is a behavior scoring model from Respan. It is listed on api14 under the decisions category, and its provider release date is 2026-09-26. Rather than generating free-form prose, it is built to read a conversation span and return, for each plain-language behavior you define, the probability that the behavior is present. That makes it a scoring component in a pipeline rather than a general assistant, and it is aimed at teams that need a repeatable numeric signal about what is happening inside a conversation.

On the input side, Span-01 accepts text. On the output side, it produces decisions, meaning the result is a set of behavior probabilities rather than a chat reply. The context length is not published in the model metadata, so you should not assume a specific token budget and should instead validate the span sizes you plan to send. The same applies to output size: the maximum output is listed as 0, which reflects that this model returns scores rather than generated text, so there is no meaningful completion length to plan around. The model exposes no supported request parameters beyond the basics of the call itself.

The practical uses follow directly from that design. You can score a support conversation for behaviors such as whether the agent acknowledged the customer's problem, whether a refund was promised, or whether the customer expressed frustration. You can run the same span against several behavior definitions at once and use the resulting probabilities as features in routing, QA sampling, escalation, or compliance review. Because the behaviors are described in plain language, you can add or change a definition without retraining anything, which suits teams whose policies shift over time. It also fits offline analysis, where you score historical spans in bulk and look for trends.

You call it through api14's OpenAI-compatible API using the model id "respan/span-01". Because the interface is OpenAI-compatible, you can point an existing OpenAI client at api14's base URL and set the model field to that id, with your api14 key for authentication. The model lists no supported parameters, so keep requests minimal and avoid assuming that sampling controls such as temperature or top_p will change the output. Treat the response as a decision payload and parse the behavior probabilities from it rather than expecting message content.

Span-01 is a narrow tool, so it is worth being clear about when it is not the right choice. It only takes text, so anything involving images, audio, or files is out of scope. It does not produce conversational text, so if you need an assistant to draft, summarize, or answer, a general chat model fits better. And since the context length is not published, very long transcripts may need to be split into spans before scoring.