Auto Router

Auto Router is a model from OpenRouter, identified on api14 as openrouter/auto. Rather than being a single fixed model, it is a routing layer: OpenRouter describes it as automatically selecting the best model for your prompt, powered by the wisdom of the market, routing based on what the community collectively spends. It sits in the image main category and also carries the text category, so it is intended for work that spans both. The provider release date is 2023-11-08.

On the input side, Auto Router accepts text, image, audio, file and video, which is a broad set of modalities for a single endpoint. It produces text and image output. The context window is 2,000,000 tokens, large enough to hold very long documents or mixed media payloads in one request. Because routing happens behind the scenes, the concrete model that serves a given call can vary with the prompt, but the interface you code against stays the same.

The practical uses follow from that combination. Long context plus file and text input suits document analysis, summarisation and question answering over large corpora. Image and video input with text and image output suits captioning, visual inspection, and generation or editing tasks where you want a picture back. Audio input extends the same pipeline to transcription-adjacent or spoken-content workflows. Tool and structured output support makes it reasonable for agent style applications that need to call functions or return JSON shaped responses, and the web search option helps when a prompt depends on current information.

You call it through api14's OpenAI-compatible API using the model id openrouter/auto, so existing OpenAI SDKs and clients work with only a base URL and key change. 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. Sampling controls, penalties, seeds, stop sequences, log probabilities and reasoning controls are all available, alongside tools and structured outputs for programmatic use.

One thing to weigh is that routing decisions are made for you, so if you need a specific model's exact behaviour, determinism across calls, or a fixed output ceiling, a named model may fit better. The maxOutput value is not published here, so plan around that uncertainty. Auto Router is best treated as a convenience endpoint for mixed modality, long context work where you would rather not pick a model per request.