The First Purpose-Built Industrial AI Model: Powering Squint’s Industrial Intelligence Platform

Today, I’m excited to announce Squint’s new, purpose-built AI model for industrials. We’re already powering global leaders like Continental, Kraft Heinz, and The Raymond Corporation. This release represents our next chapter for our customers.

The model is the critical innovation in our new AI system, which brings together tribal knowledge, data from IT systems, and standard work protocols. The system makes all that context available to industrial agents, so AI can solve real problems and create tangible impact for global manufacturers and field services providers.

We’re rolling it out now.

For too long, in manufacturing and field services, knowledge has been fragmented across people, processes, and systems—trapped in peoples’ heads, sitting in dusty binders, or buried in cumbersome Sharepoints. This has made it incredibly hard for operators to access the information they need when they need it, and impossible for any AI system to access the context required to deliver meaningful productivity gains. 

So we built the solution: to capture industrial knowledge and activate it. Even knowhow that’s never been written down. It’s time to actually deliver on the promise of AI in industrials.

The industrial context layer

With this model, Squint can create a knowledge graph that serves as each customer’s context layer for apps, agents, and workflows. Squint will fit like a glove for any organization: from creating missing work instructions, to running scrap mitigation workflows, to ensuring faults are resolved in real-time.

In order to build this complete context layer, we needed more than a fine-tuned model. We had to capture and vectorize tribal knowledge: operators' memories and experiences. That’s the real gold mine. Almost no interesting things are already documented on factory floors. Squint’s AI-powered knowledge capture can ingest that insight, and feed it into the knowledge graph for a richer, more complete context layer than anything else on the market. 

The model

The obvious initial path was to point an off-the-shelf model at a customer's files. But frontier models can’t deliver the accuracy and completeness that high-stakes manufacturing and field services require. 

So, we post-trained an open-weight, 2-billion parameter model on an industrial ontology to outperform the closed-weight frontier models for industrial applications.

We tested this on a benchmark of the hardest questions customers have asked us; questions that veteran operators could only answer after decades of hands-on experience. In these evaluations, our system scored 78% accuracy versus Claude Code's 53% and OpenAI’s 46%—a 25+ point difference.

The systems we tested alongside Squint are: OpenAI Assistants (their file search API), Vector RAG (a classic RAG setup using semantic similarity over embeddings), and Claude Code (Anthropic's agent reading a document folder).

Squint’s system isn’t only the highest-performing; it’s the approach best-suited for continuous learning. Because our model for creating these knowledge graphs is so efficient, we can update them to reflect up-to-the-minute learnings from the work that’s happening inside Squint: every procedure run, task completed, and query sent to Assistant. Squint’s model can create knowledge graphs 1000x faster than it would take for GPT-5 to create the same context layer. 

Fully-deployed for customers

Every industrial organization is unique, and our approach ensures every customer can have a scalable, custom solution. Our platform approach makes continuous learning seamless: work happens inside Squint and feeds back into the knowledge graph, which gets richer and deeper. 

This new foundation makes our Connected Worker features even more capable, right out of the box. And it creates infinite potential for custom workflows and agentic solutions for industrial work's toughest problems.

To unlock that potential, our team at Squint will be partnering directly with industrial organizations to learn about your biggest bottlenecks, and use our AI to design and execute end-to-end solutions. Squint is now entirely configurable on top of your custom context—your business priorities, solved in the flow of how your work actually gets done.

We’re just getting started.