Onton Releases Ontology 1: A Neurosymbolic Search Model That is 2.7x More Accurate than E-commerce Giants
Onton launches Ontology 1, a neurosymbolic e-commerce search model backed by a custom graph database, outperforming Amazon and Google Shopping in recent benchmarks.
Stock photo for illustration only, not from the actual event
- Onton has launched Ontology 1 live for end users directly on Onton.com.
- Evaluated using the Subtext-Decor-90 benchmark against Amazon and Google Shopping.
- Powered by Ograph, a custom high-performance graph database architecture.
- Currently available exclusively through case-by-case partnerships with no public API.
Onton has officially released Ontology 1, a neurosymbolic search model that is now live for end users at Onton.com. For development teams building on the agentic web, partner access is granted on a case-by-case basis. At present, there is no public API, pricing tier, or open checkpoint available for the model itself, meaning adoption requires a direct partnership rather than a standard pip install.
Traditional e-commerce interfaces have remained largely unchanged for nearly 30 years, relying on the assumption that user intent maps strictly to basic categories and attributes like size, price, material, and brand. These legacy systems lack filters for nuanced concepts such as pet-friendly items or furniture dimensions that precisely match a specific room layout.
To demonstrate performance, Onton released the Subtext-Decor-90 benchmark along with its code and data. Three multimodal judges—Claude Opus 4.8, Gemini 3.1 Pro, and GPT-5.5—scored the top 10 visible result cards returned by Onton, Amazon, and Google Shopping across 90 text queries. P@10 results were averaged across the judges with 95% confidence intervals derived from 10,000 bootstrap resamples.
Stock photo for illustration only, not from the actual event
The results show Onton leading with a score of 0.630, followed by Google Shopping at 0.543, and Amazon at 0.469. Onton won 52 queries outright, compared to 19 for Google and 16 for Amazon. These figures sum to 87 because Ontology returned fewer than 10 results on three queries, with empty slots scored as non-relevant. Excluding those specific slots adjusts the scores to 0.665 for Onton, 0.549 for Google, and 0.459 for Amazon.
The integration of neurosymbolic AI into retail search engines represents a significant paradigm shift away from rigid keyword matching. By bridging deep learning with symbolic reasoning, models like Ontology 1 can interpret complex semantic intents and relational contexts, addressing limitations that have persisted in digital commerce discovery for decades.
Under the hood, Ontology 1's knowledge graph runs on Ograph, a custom graph database. Onton reports that a single Ograph core outperforms SuiteSparse:GraphBLAS running on 14 cores, achieving roughly 100x the throughput per core, while its GPU implementation runs 43x faster than the CPU variant with early tuning peaks reaching up to 1000x.
Source: MarkTechPost
Found something wrong in this article? Report an issue with this article
Comments
Leave a Comment