ALEX UPDATE 
ALEX 13.0 AVM General Availability - Just Released Best AVM Model Available
12 Training Sessions  |  116-Tier Cascade  |  3-Stage Stacked Ensemble  |  Supersedes ALEX 12.8 | Now available to all ALEX agents & clients

Valuation based on what sellers net.

Net sale price is the ALEX standard — sale price LESS buyer closing costs paid by seller, LESS repairs required by Buyer during the Option Period and even after closing. 

Final ALEX 13.0 valuations normally fall within 2.58% blended MAPE and 2.18% blended MdAPE across all single family segments in ABOR's primary service area spanning 18 Counties. Data: Closed sales July 15-Aug 14, 2026. 2,582 transaction reported. 8 eliminated due to probable error. So number (N) = 2,574 transactions.

AAEX 13.0 replaces the prior comparable selector with a time-normalized, three-stage stacked ensemble. The selection cascade expands from 66 tiers to 116, training sessions double to 12 and the number of incorporated statistical models triples.

Every ABOR listing since 2016 is included in our data lake. A data lake sort of like a database but stores and can research data in any format. Commonly referred to as unstructured data. A good example is ABOR listing information data, County databases, images, documents and retrieval augmented generated (RAG) data from the web.

ALEX captures all listings regardless of status (Active, Active Under Contract, Pending, Closed, Hold, Withdrawn, Expired and "Reduced (calculated by ALEX.)" Data is updated every 15 minutes and refreshed monthly. 

Data includes both standard RESO and non-RESO API fields. Non-RESO fields are not captured by other AVMs. Examples of Non-RESO fields are Closing Costs Paid by Seller and Repairs paid by Seller at Buyer Request.

Comparable selection runs through a cascade of 116 tiers of agentic and subagent AI, preferring the same subdivision, builder, and neighborhood section where the data supports it. 

For each target and comparable listing, each picture is analyzed through largest available models including: Qwen2-VL (72 billion parameter size) or LLaVA (34 billion parameter size) served through Ollama (405 billion parameter size.) Subject to change almost daily!

Published statistics accompany every detailed valuation. Forecast Standard Deviation (FSD), Confidence Interval (CI), Root Mean Squared Logarithmic Error (RMSLE or back-test error) are always provided for each detailed valuation.

National AVM (Zillow, Core Logic now Cotality, and HouseCanary unless not available for a specific property) values are used for price range benchmarks. 

Current and prior year County Tax valuations are used for pricing and trend reference. If an API is available, we use it now or will add it. If not, we download data, query in real time and cross validate.

"ALEX 13.0 was built to outperform all valuation tools. It is MIT dissertation committee presentation level precision and documentation, not a marketing department. We publish statistics and define all statistical models used in each valuation" 
Diane Hart Alexander, MBA, MHA,
Designated Broker, 
ALEX Real Estate 
Chair & CEO, 
ALEX Companies 

Headquartered in Georgetown, Texas in the Austin metro area. 

ALEX Real Estate is organized around clients not teams, property type, size or location. So, ALEX teams, agents and associate brokers can manage any client transaction, any type, any size, almost anywhere. Limited only by U.S. State Department guidelines.

ALEX Real Estate is part of the ALEX companies: ALEX Real Estate, ALEX Intelligence, ALEX Advisory, ALEX Institute and ALEX Cares (Philanthropy.)

Want more information? 

Want to discuss:
AI & real estate?
AI Sovereignty vs Cloud Computing?
AI & Quantum Computing?
AI tools & their evolution/revolutions (today and tomorrow)
Real estate: in general, specific micro market or region?
ALEX?
Something completely different?
Contact George.