Elorian: Glimpse

Elorian: Glimpse is a text model from the provider Elorian, listed on api14 under the main category of text. It is described as a multimodal visual reasoning model built for visual understanding and complex visual reasoning in the physical world, and it reasons directly over images and video rather than relying on a separate description step. If you are choosing a model for work that involves interpreting what is happening in visual media, this is the model Elorian positions for that job.

The model accepts text, image and video as inputs and returns text. Its context window is 262144 tokens, and the maximum output is 256000 tokens, so long inputs and long generations are both possible within the same request. Because the input side is multimodal while the output side is text only, the model is suited to tasks where visual material goes in and a written answer comes out, such as describing, comparing or reasoning about what appears in an image or a video.

Practical uses follow from that combination. You can send a video and ask for a summary of the events in it, or send a set of images and ask for a comparison across them. Visual question answering, inspection of physical scenes, and reasoning about spatial or physical relationships in footage are all natural fits. The large context window also means you can supply substantial visual material or lengthy accompanying text in one call rather than splitting the work across many requests.

You call the model through api14's OpenAI-compatible API using the model id "elorian/glimpse". The request parameters it supports are frequency_penalty, include_reasoning, logit_bias, logprobs, max_tokens, min_p, presence_penalty, reasoning, repetition_penalty, response_format, stop, structured_outputs, temperature, tool_choice, tools, top_k, top_logprobs and top_p. Tool calling and structured outputs are both available, so you can combine visual reasoning with function calls or with responses constrained to a schema, and the reasoning and include_reasoning parameters let you control whether reasoning content is surfaced.

Keep the output modality in mind when planning: the model returns text, so anything you need in another form has to be produced downstream. If your task is purely textual and involves no images or video, the visual reasoning focus of this model may be more than you need, and a text-only model could be a simpler fit. For work that turns on understanding what is visible in images or video, Glimpse is the appropriate choice among the models available here.