d1.2-ultra overview
Decision example
null
| Question | Result |
|---|---|
| Damaged item | P(yes): 0.4185 |
| Team | billing (confidence 0.0480) |
| Urgency | 0.7742 / 2 (confidence 0.1519) |
Actual model response from Nov 17, 2026.
View request and response
Request
{
"model": "liquid/d1.2-ultra",
"questions": {
"damaged_item": {
"type": "noul",
"instructions": "Does the customer report an item that arrived damaged?"
},
"team": {
"type": "choice",
"instructions": "Which team should handle this request?",
"criteria": {
"returns": "Damaged products, replacements and returns",
"billing": "Payments, invoices and charges",
"technical": "App errors and integrations"
}
},
"urgency": {
"type": "score",
"instructions": "Rate how urgently the request needs attention.",
"criteria": [
"Routine follow-up",
"Customer is blocked",
"Immediate safety risk"
]
}
}
}Response
{
"model": "d1.2-ultra",
"answers": {
"damaged_item": {
"type": "noul",
"noul": 0.4185247320898998
},
"team": {
"type": "choice",
"choice": "billing",
"probabilities": {
"returns": 0.31732512519144157,
"billing": 0.36534974961711675,
"technical": 0.31732512519144157
},
"confidence": 0.04802462442567512
},
"urgency": {
"type": "score",
"score": 0.7742374329462929,
"legend": {
"0": "Routine follow-up",
"1": "Customer is blocked",
"2": "Immediate safety risk"
},
"probabilities": {
"0": 0.39559778082236674,
"1": 0.4345670054089734,
"2": 0.16983521376865973
},
"confidence": 0.15185050811346024
}
},
"usage": {
"input_tokens": 128,
"output_tokens": 0
}
}d1.2-ultra by Liquid AI
Liquid AI describes d1.2-ultra as its most capable decision model. The new version adds audio input and five additional task types.
d1.2-ultra specs
- Model ID
liquid/d1.2-ultra- Developer
- Liquid AI
- Context window
- 65,536 tokens
- Input
- Image, Text
- Output
- Text
- Endpoint
POST https://api.aimlapi.com/v1/decisions
Many questions about one input
A request carries a map of 1 to 128 questions, each under its own key. Every key comes back in the answers with a result of the type you asked for. Liquid AI states that the model processes all questions in a single forward pass and generates zero output tokens.
Put every decision you need about one input into the same request. A moderation flag, a topic label and a quality score for the same message can share one call.
Pick the question type for the decision
A yes/no question returns the probability of yes between 0 and 1. You can add a description of what true and false mean for that question, as text or as a JSON object or array. Start with this type for moderation gates.
A choice question takes 2 to 255 options. It returns the selected key, a calibrated confidence and a probability for each option. Turn on the multiple setting when more than one option can apply. Give each option a description, or leave it empty when the key says it all.
Choose it for harder decisions
Liquid AI recommends d1.2-ultra for harder classification, routing, scoring, extraction and other decision tasks. It also says the model supports audio-based decisions such as identifying a spoken language or detecting a siren.
The inputs listed for this endpoint are image and text. Start with a text or image workload and test your hardest cases there.
Use cases
Flag unsafe or off-policy user messages with yes/no moderation questions and a probability threshold.
Route support tickets to one of several queues with a single choice question.
Grade another model's answers on an ordered scale as an LLM-as-judge check.
Inspect product photos against several yes/no questions in one request.
Capabilities
- Vision
d1.2-ultra API pricing
| Usage | Price | Unit |
|---|---|---|
| Input | $0.042 | 1M tokens |
| Output | $0 | 1M tokens |
Prices from the AI/ML API price list, checked on Nov 17, 2026.
How to use the d1.2-ultra API
View docs (opens in a new tab)Build with your agent
Give your agent the model details, code and documentation.
1Get your API key
Create API keyexport AIMLAPI_KEY="8f2b4c7d9a1e6b3f…"2Make your first request
curl 'https://api.aimlapi.com/v1/decisions' \
-H "Authorization: Bearer $AIMLAPI_KEY" \
-H 'Content-Type: application/json' \
-d '{
"model": "liquid/d1.2-ultra",
"questions": {
"damaged_item": {
"type": "noul",
"instructions": "Does the customer report an item that arrived damaged?"
},
"team": {
"type": "choice",
"instructions": "Which team should handle this request?",
"criteria": {
"returns": "Damaged products, replacements and returns",
"billing": "Payments, invoices and charges",
"technical": "App errors and integrations"
}
},
"urgency": {
"type": "score",
"instructions": "Rate how urgently the request needs attention.",
"criteria": [
"Routine follow-up",
"Customer is blocked",
"Immediate safety risk"
]
}
}
}'import os
import requests
headers = {
"Authorization": f"Bearer {os.environ['AIMLAPI_KEY']}",
"Content-Type": "application/json",
}
payload = {
"model": "liquid/d1.2-ultra",
"questions": {
"damaged_item": {
"type": "noul",
"instructions": "Does the customer report an item that arrived damaged?"
},
"team": {
"type": "choice",
"instructions": "Which team should handle this request?",
"criteria": {
"returns": "Damaged products, replacements and returns",
"billing": "Payments, invoices and charges",
"technical": "App errors and integrations"
}
},
"urgency": {
"type": "score",
"instructions": "Rate how urgently the request needs attention.",
"criteria": [
"Routine follow-up",
"Customer is blocked",
"Immediate safety risk"
]
}
}
}
response = requests.post("https://api.aimlapi.com/v1/decisions", headers=headers, json=payload, timeout=120)
response.raise_for_status()
print(response.json())const response = await fetch("https://api.aimlapi.com/v1/decisions", {
method: "POST",
headers: {
Authorization: `Bearer ${process.env.AIMLAPI_KEY}`,
"Content-Type": "application/json",
},
body: JSON.stringify({
"model": "liquid/d1.2-ultra",
"questions": {
"damaged_item": {
"type": "noul",
"instructions": "Does the customer report an item that arrived damaged?"
},
"team": {
"type": "choice",
"instructions": "Which team should handle this request?",
"criteria": {
"returns": "Damaged products, replacements and returns",
"billing": "Payments, invoices and charges",
"technical": "App errors and integrations"
}
},
"urgency": {
"type": "score",
"instructions": "Rate how urgently the request needs attention.",
"criteria": [
"Routine follow-up",
"Customer is blocked",
"Immediate safety risk"
]
}
}
}),
});
if (!response.ok) throw new Error(`HTTP ${response.status}`);
console.log(await response.json());d1.2-ultra FAQ
How much does d1.2-ultra cost?
d1.2-ultra is billed at $0.042 / 1M tokens for input. There are no output-token charges.
How many questions can one request hold?
A request holds 1 to 128 questions. Each question has its own key, and every key comes back in the answers with a typed result.
Can a choice question return more than one option?
Turn on the multiple setting and the model selects every option that applies. A choice question accepts 2 to 255 options, and the response includes a probability for each.
How do I read a score answer?
The score is the expected level on your ordered scale and may be fractional. The response also returns a confidence, the legend you supplied and a probability for each level.
What does a yes/no question return?
It returns the probability that the answer is yes, a value between 0 and 1. You can add descriptions of what true and false mean for that question.
What is the d1.2-ultra model ID?
Use liquid/d1.2-ultra as the model parameter. Requests go to https://api.aimlapi.com/v1/decisions with your AI/ML API key.