Questel Launches QaECTER AI for Patent Search
Questel, a world leader in intellectual property solutions, including AI-assisted patent search, translation, and filing, has announced the release of QaECTER — a new AI model for patent search. Developed and optimised by Questel’s AI Lab scientists, QaECTER reaches state-of-the-art capabilities in semantic patent retrieval. Customers using Orbit Intelligence, its AI agent Sophia, and other products with the semantic search feature will substantially benefit from QaECTER.
To evaluate the performance of the patent search model, Questel’s AI Lab developed Sophia-Bench, a systematic, large-scale benchmark comprising 10,000 patent queries and 75,000 corpus patents, stratified across 10 years.
In exacting tests using Sophia-Bench and an independent external benchmark, QaECTER outperforms all competing general-purpose and patent-specific retrieval models across every query type, technology domain, and jurisdiction evaluated. These gains stem from novel training methodologies developed by the Questel AI Lab, combining citation-driven supervision with multi-view self-alignment on Questel’s proprietary patent data.
Kim Gerdes — Questel’s AI Lab Director, Co-Founder and CTO of qatent (part of Questel since 2024), and a Professor of Computer Science at Paris-Saclay University — co-authored both the paper and the report. He notes: “Our exceptional team of AI and natural language processing experts collaborates with top academic institutions. We are dedicated to advancing AI in the IP field with state-of-the-art models, such as QaECTER, and evaluation tools, such as Sophia-Bench. These new developments place our team at the very forefront of modern deep learning for information retrieval.”
About QaECTER
The QaECTER model places patent search practitioners at the centre of its training methodology. It learns from how examiners and professionals establish relevance between documents, through citation relationships and multiple complementary views of each patent. The model performs consistently across diverse inputs, including invention disclosures and problem statements, as well as claims, abstracts, and technical or figure descriptions.
This supports R&D engineers, patent attorneys, examiners, and strategic searchers alike. Within Sophia Search, QaECTER is complemented by an integrated translation layer for multilingual queries, advanced reranking, and additional capabilities designed for professional patent workflows. Orbit Intelligence customers who perform semantic patent searches via Sophia Search have observed strong performance across a wide range of technical domains.
About Questel
Questel is an end-to-end intellectual property solutions provider serving 20,000 organisations in more than 30 countries to optimise the management of their IP asset portfolios.