Auto Router (Beta)

Auto Router (Beta) is a model from OpenRouter, listed under the model id openrouter/auto-beta. It sits in the image category while also covering text, which reflects the fact that it accepts more than one kind of input and can return more than one kind of output. OpenRouter describes it as the experimental version of its Auto Router, a place where new improvements are tested before they reach the general purpose router. The provider notes that you should expect beta behaviour, so treat it as a moving target rather than a fixed, frozen model.

On the input side, the model accepts text, image, audio, file and video, so you can send a prompt alongside rich media in a single request. On the output side it produces text and image, which means a call can return written content, generated images, or both depending on what you ask for. The context window is 2,000,000 tokens, which is large enough to hold long documents, extended transcripts or sizeable media collections in one request. The maximum output length is not specified in the model details.

Those capabilities point to a few practical patterns. Long context plus file and video input suits document analysis, media review and summarisation work where you want to pass a lot of source material at once. Because the model can emit images as well as text, it fits mixed tasks such as generating a written explanation together with a visual, or producing image output from a text and image prompt. Audio input extends this to transcription-adjacent or spoken-instruction workflows, and video input covers cases where the model needs to reason about moving footage rather than a still frame.

You call it through api14's OpenAI-compatible API using the model id openrouter/auto-beta, so existing OpenAI client libraries work with only the base URL and key changed. The supported request parameters are frequency_penalty, include_reasoning, logit_bias, logprobs, max_tokens, min_p, prediction, presence_penalty, reasoning, reasoning_effort, repetition_penalty, response_format, seed, stop, structured_outputs, temperature, tool_choice, tools, top_a, top_k, top_logprobs, top_p and web_search_options. Tool calling, structured outputs and web search options are available, which helps when you need the model to call functions or return data in a fixed shape.

The main caveat is the beta label. OpenRouter states plainly that this is the experimental version of the Auto Router, so behaviour may change as improvements are tested. If you need stable, predictable output for production, a non-experimental model may fit better. If you want early access to router improvements and can tolerate change, this model is the one to try.