Aimlapi · Chat
Dynamic Router
AimlapiChat
A router you configure in the request. Describe up to ten buckets in your own words (`when`), each with the models to try in order, and name the decision model: it picks the bucket whose description fits each request — its text and its images — and the first model of that bucket that can take the request answers, falling back to the next one in your order. One `router` serves chat completions, the Responses API (`/v1/responses`), the Messages API (`/v1/messages`), image generation and editing, and video generation: list models of every kind in a bucket, and each request takes those of its own kind — on Responses and Messages, the models served there. A bucket with no models (`models: []`) ends the request at the decision: no model is called, the answer is empty (an empty message, no images, a video generation with no video), and `meta.router` gives your code the bucket to act on (`routed: false`) — only the judgement is charged. One bucket makes a plain fallback chain, with no judgement. Without a `router` object it routes a conversation like the Jev Router — light: GPT-6 Luna, Claude Haiku 5.5, DeepSeek V4.1 Flash, Gemini Flash-Lite; standard: GPT-6.1 Sol, Claude Sonnet 5.5; heavy: Claude Opus 5.5, GPT-6 Astra; on Responses and Messages, those of them served there — and a `router` that leaves a setting out keeps its default; image and video requests need buckets of your own. You pay the rate of the model that answers, named in `meta.model`, plus the judgement: one decision at its catalogue price, its images included, listed as its own request in your usage. `meta.router` says which bucket the request landed in, why, what the decision model chose and how sure it was, and how many images it was shown.
Evidence and sources
Released 30 Sep 2026
Sources: AI/ML API data, llminfo researchPrimary data from AI/ML API
from llminfo research: release date
Input & output
from the model's typeText
Text
Also known as: aimlapi/dynamic-router
Like this one. Tap or hover any dotted value, label or icon to see where it comes from and how it is measured. open details when tapped or hovered. How this data works