We use cookies. Cookies and browser storage keep you signed in and remember your preferences. With your permission we also measure how the site is used. You can any time. See our cookie policy.
~liquid/d1-ultra-latestRuns on your balance at the prices above. New accounts start with free usage.
Settings
LiquidAI: d1-ultra Latest is a model from LiquidAI, and it sits in the decisions category. It is a redirecting alias: the model always points to the latest model in the d1-ultra family, so the underlying version can change over time while the model id stays the same. That behaviour is worth knowing before you build against it, because pinning to a moving target is different from pinning to a fixed snapshot.
The model accepts text and images as input, so it can work with prompts that combine written instructions and visual material. Its output modality is decisions rather than free-form prose, which means you should expect structured choice output rather than a chat-style answer. The context window is 32768 tokens, and the maximum output is 29491 tokens. Those two numbers together mean a request can carry a substantial amount of input and still leave room for a large decision output.
The practical uses follow from that shape. Because it takes images as well as text and returns decisions, it suits classification and routing tasks where the answer is a label, a category or a choice among options. Examples include triaging incoming requests, sorting documents or screenshots into buckets, and picking a branch in an automated workflow. The image input opens up cases where the deciding signal is visual, such as judging a photo or a rendered interface, while the text input carries the rules or context for the decision. The large output allowance is useful when a single call needs to return many decisions at once rather than one.
You call it through api14's OpenAI-compatible API using the model id "~liquid/d1-ultra-latest". Because the endpoint follows the OpenAI chat completions shape, you send your messages in the usual way and read the response from the standard fields. The model supports no additional request parameters beyond the common ones, so there is no separate tuning surface to configure; keep your request to the standard fields and put the task instructions in the prompt. Since the id is a redirect, the same request may be served by a different underlying d1-ultra version later, so test again if behaviour shifts.
If your task needs generated text, conversation or explanations rather than a decision, a model with a text output modality will fit better. The 32768 token context is also a real ceiling: long documents, large image sets or extended histories may need chunking or a model with a longer window. And if you need reproducible behaviour across time, the redirecting nature of this id is a reason to check whether a fixed version is available instead.
import OpenAI from "openai";
const client = new OpenAI({
baseURL: "https://api14.com/api/v1",
apiKey: process.env.API14_API_KEY,
});
const result = await client.post("/decisions", { body: {
"model": "~liquid/d1-ultra-latest",
"state": "Customer: I was charged twice for my order and want my money back.",
"questions": {
"team": {
"type": "choice",
"instructions": "Which team should handle this message?",
"criteria": {
"billing": "Payments, refunds, invoices",
"support": "Product questions and bugs",
"sales": "Pricing and new purchases"
}
},
"urgent": {
"type": "noul",
"instructions": "Does this need a reply today?",
"criteria": {
"true": "Money, outage or legal risk",
"false": "Can wait"
}
}
}
} });
console.log(result.answers);
This community has no topics yet. Open the first one about LiquidAI: d1-ultra Latest.