---
title: 'NEW Liquid AI Models are CRAZY!'
source: 'https://youtube.com/watch?v=-STC6XfFpRI'
video_id: '-STC6XfFpRI'
date: 2026-11-18
duration_sec: 492
channel: 'Julian Goldie SEO'
---

# NEW Liquid AI Models are CRAZY!

> Source: [NEW Liquid AI Models are CRAZY!](https://youtube.com/watch?v=-STC6XfFpRI)

## Summary

Liquid AI has released two new decision-making models, D 1.2 and D 1.2 Ultra, which output structured, machine-readable data instead of plain text. These models are designed for business automation, offering five new capabilities: value extraction, span detection, object localization, numerical estimation, and multi-select decisions. The video demonstrates practical workflows for lead qualification and support routing, positioning these models as a significant upgrade over traditional chatbots.

### Key Points

- **Introduction to Liquid AI D 1.2 and D 1.2 Ultra** [00:03] — Liquid AI released two new models, D 1.2 and D 1.2 Ultra, which make structured decisions from text, images, and audio, returning machine-readable outputs for automations.
- **Key Differentiator: Structured Outputs** [00:31] — Unlike most AI tools that provide paragraphs of text, these models return clean, organized, machine-readable answers with specific values, classifications, and outputs that workflows can act on immediately.
- **D 1.2 Ultra Ranking** [01:01] — D 1.2 Ultra hit number one on the Artificial Analysis Decision Index, a benchmark for decision-making AI models.
- **Capability 1: Value Extraction** [01:43] — The model reads documents (text, PDF, images) and extracts specific values like names, numbers, and dates, delivering clean structured data for direct use in CRMs or other systems.
- **Capability 2: Span Detection** [02:13] — This feature finds every passage in a document matching a user-defined category, such as all phone numbers or questions, eliminating manual searching.
- **Capability 3: Object Localization** [02:41] — For images, the model locates specific objects (e.g., logos, signatures, damaged items) and returns bounding box coordinates for use in automation.
- **Capability 4: Numerical Estimation** [04:09] — The model estimates counts from images (e.g., number of people in a photo) and provides a confidence score alongside the estimate.
- **Capability 5: Multi-Select Decisions** [04:27] — Unlike traditional single-label classification, this allows classifying a single input under multiple categories simultaneously, e.g., a support message can be flagged as both a delivery issue and a billing question.
- **Workflow Example: Lead Qualification** [05:13] — D 1.2 reads form submissions, extracts details, and classifies leads into categories (hot prospect, needs more info, etc.) with confidence scores, enabling automated routing and follow-up.
- **The Opportunity Window** [06:28] — Most businesses still use AI as a chatbot; adopting decision models now provides a competitive head start before these capabilities become mainstream.

## Transcript

the end of this video, you're going to see exactly why business owners are paying close attention to this one. Liquid AI just released two new models. Liquid AI just released two new models. They're called D 1.2 and D 1.2 - Ultra.
And here's what makes them different from everything else out there right now. These models don't just answer your questions. They make structured decisions from text, from images, from audio.
And they return those decisions in a format that your automations can actually use. Think about that for a second. Most AI tools give you a paragraph of text and you have to figure out what to do with it. These models
give you back clean, organized, machine-readable answers, specific values, specific classifications, specific outputs your workflows can act on immediately. That's a completely different use case. And D 1.2 Ultra just
hit number one on the Artificial Analysis Decision Index. We'll get into what that means, but first, let me show you what these models actually do because this is the part most people are sleeping on. Liquid AI has been building
a different kind of AI company from day one. They're not trying to build another chatbot. They're building models designed for specific high-value tasks in real business environments. The D 1.2 release is the second generation of
their decision model line. And this upgrade added five brand new capabilities that weren't there before. Let me walk you through them because each one opens up a completely different automation opportunity.
The first one is called value extraction. The model reads a document, extraction. The model reads a document, a text file, a PDF, an image of a form, and it pulls out specific values, names, numbers, dates. It doesn't describe the
document to you. It hands you the actual data, clean in a structured format. So if you get a form submission through your website, you don't manually copy the information into your CRM. model reads it and extracts it for you
automatically. The second capability is span detection. This one finds every passage inside a document that matches a certain category you define. So if you tell it find every phone number, it goes through the whole
phone number, it goes through the whole document and returns every single one. If you tell it find every question someone has asked, it pulls out just the questions. No manual reading, no control F hunting. It reads the whole thing and
returns exactly what you asked for. Third, object localization. This is for images. The model can look at a photo and locate specific objects inside it using bounding boxes. Find the logo, find the signature, find the damaged
item. It tells you where it is in the image with coordinates you can actually use. Quick pause before I show you the next part because if you want to next part because if you want to actually build with Liquid AI D 1.2 and
not just watch someone else do it, this matters. Inside the AI Profit boardroom, matters. Inside the AI Profit boardroom, we've built a full 30-day D 1.2 roadmap. Day one, you get your API key set up and run your first value extraction on a
real document. Day two, you connect it to Make or N8N and build your first automated input pipeline. Day three, you add confidence thresholds so your workflow knows when to process automatically and when to flag for human
review. By the end of the first week, you've got a live decision model running in your business. We run four coaching calls every week where you can bring your exact D 1.2 setup and get it reviewed live. And there are already
members inside running D 1.2 workflows for lead qualification, support routing, and document processing right now. Link in the description or go to AIprofitboardroom.com. Fourth, numerical estimation. The model
looks at an image and estimates a count. many people are in this event photo? How picture? And it doesn't just give you a number, confident it is. Fifth, multi-select decisions. This is a
big one. Most AI classification forces you to pick one label, but in real life one message can be about multiple things at once. With multi-select the model can classify a single input under multiple categories
at the same time. A support message can be a delivery issue and a billing question and an urgent case all at once and the model will flag all three. That last one alone changes how customer support automation works. Now let's talk
about the actual workflows and every single one of these is based on a real use case for growing the AI profit boardroom so you can see exactly how decision models work in a real business context. Workflow one, automating lead
qualification for the AI profit boardroom. What this does, when someone fills out the AI profit boardroom interest form, D 1.2 reads the submission, extracts their details and classifies the lead so the right
follow-up goes to the right person without anyone manually reading the Prompt to use in your automation. You are a lead qualification assistant for the AI profit boardroom. Read the following form submission. Extract full
name, business type, current tool they mentioned and urgency level. Classify the lead into one or more categories, hot prospect, needs more categories, hot prospect, needs more info, already a member, not a fit.
confidence scores for each classification. Result, every new lead gets instantly classified, routed, and followed up with zero manual sorting. Hot prospects get a reply fast. The businesses that move on
tools like this now are going to be ahead. The ones that wait until these capabilities are mainstream will be catching up. And right now, most people haven't even heard of D 1.2. That's your window. Here's the thing.
There's a window right now where most business owners haven't figured out how to use decision models like D 1.2 in their workflows. Most people are still using AI as a chatbot. They're not building structured automation with it.
That window won't stay open forever. The businesses that move on this now are going to have a real head start. Inside the AI Profit boardroom, we've got daily tutorials walking you through exactly how to connect D 1.2 to your automation
how to connect D 1.2 to your automation stack, step by step, no guesswork. Four live coaching calls every week where you can get your specific workflows reviewed and your questions answered in real time. A prompt library built around
decision model automation, lead qualification prompts, support routing prompts, document extraction prompts, all ready to use. And a member map so you can find and connect with other people near you who are building with
the same tools. Link in the description or go to aiprofitboardroom.com. And if you want the full process notes from this video, the prompts, the workflow blueprints, the step-by-step
SOPs, come join the AI Success Lab. It's our free community. Links are in the comments and the description. You'll get everything from this episode in there, plus access to over 100 AI use cases just like this one.
just like this one. Liquid AI D 1.2 and D 1.2 Ultra are live. The number one ranked decision AI is available through an API right now. Five new capabilities, text, images, audio, all in one model. And practical
workflows you can start building today. Go check it out. Links are below.
