Blog

Our insights, engineering notes, and stories from applying AI to drawings, specs, and documents.

AI for Drawings

ChatGPT or Claude: Which AI Is Best at Reading Architectural Drawings?

As of October 2026, GPT-6 Astra has the highest overall score in COXIT’s CaseVBench evaluation of architectural drawings, at 92%. Claude Opus 5.5 scored 79% and…

October 14, 2026

//Category

AI for Drawings

See all 10 articles

Top 8 AI benchmarks for drawings: CaseVBench, Construction’s Last Exam, TakeoffBench-v1, AECV-Bench, DrawingVQA, AEC-Bench, LED-BENCH and Blueprint-Bench 2

Top 8 AI Benchmarks for Architectural and Construction Drawings in 2026

Every AI takeoff tool claims to be accurate. Some promise accuracy of up to 98%, while others advertise expert-level results. But how can you tell whether those claims reflect real performance? Until recently, there was little public data to help… Volodymyr Hresko 10/23/26

Jacob Edmond and Volodymyr Hresko on the Verify In Field podcast

Verify In Field podcast: What we learned testing AI on casework drawings?

At IWF Atlanta this summer, one piece of advice kept coming up: use Claude, Claude is the best. Our co-founder and COO, Volodymyr Hresko, kept answering with a question: is it really? He took that question, and what we found… Iryna Mykytyn 10/21/26

Gemini 4 Argon key art: the model name beside a large white number 4 on a blue gradient.

How Good Is Gemini 4 Argon at Reading Architectural Drawings?

Gemini 4 Argon scored 76% on COXIT’s CaseVBench architectural drawing benchmark, placing third out of 20 AI models. Only GPT-6 Astra, at 92%, and Claude Opus 5.5, at 79%, scored higher. What makes Argon interesting is its price. At $0.021… Volodymyr Hresko 10/19/26

An estimator scratching their head over an architectural floor plan marked with red question marks, with a calculator and a scale ruler on the desk.

Understanding F1 Scores in AI Drawing Benchmarks

F1 is the headline score in CaseVBench. A kitchen-drawing example explains precision, recall and why F1 lands nearer the weaker of the two, plus what IoU and object type change about reading a result. Volodymyr Hresko 9/15/26

//Category

See all 8 articles

We launched CaseVBench: an AI benchmark for object recognition in architectural drawings, shown over an interior elevation sheet with detected objects boxed.

COXIT Newsletter Sep 14, 2026

September 2026: Introducing CaseVBench, an Accuracy Benchmark for AI on Architectural Drawings

We ran 11 AI models over 119 hand-annotated drawing pages to see how well they find cabinets, countertops, floor plans, elevations and… Volodymyr Hresko

//Category

ML Engineering

See all 8 articles

Cartoon wizard cat with the title RAG Evaluation

Evaluating RAG using LLM-as-a-Judge Independent Evaluation, Automatic Insights Collection

The entire approach is open-sourced and is accessible as a GitHub repository. Evaluation? Why RAG output still needs checking In my latest article, “Chatting with Large PDFs,” I discussed the technical aspects of implementing a RAG for your PDFs to… Valerii Verezhynskyi 6/19/25

Architecture diagram of a chat app with MCP document search servers and OpenAI

Chatting with Large PDFs (100–500 Pages): Using RAG with OpenAI Embeddings (Local vs. API)

At COXIT, in 2025, we observe a trend when more and more potential customers ask us to implement a way to chat with massive documents or even sets of documents, search information in them, and perform an analysis dynamically. Those… Valerii Verezhynskyi 5/31/25

Grafana, Prometheus and C++ logos

How to set up monitoring of CPU and memory usage for C++ multithreaded application with Prometheus, Grafana, and Process Exporter

Have you ever wondered how to know how much CPU and memory resources is your app taking during a week? Or how to collect and visualize this data to get some useful insights? In this guide, I will show you… Iryna Mykytyn 5/23/25

How To Integrate Valgrind into GitHub Actions?

Have you ever got into the situation that after you fixed memory leaks with Valgrind you see Definitely lost again after a while? Then this CI setup may be useful for you. You can do it in 3 steps: These steps are… Iryna Mykytyn 5/14/25

Read us there

Let's collaborate

Tell us a bit about your project or challenge, and we'll get back to you shortly.

Volodymyr Hresko

Volodymyr Hresko Co-Founder & COO

Reach out directly

[email protected]