Regional deep-tech company delivers greater enterprise AI accuracy than OpenAI, with a fifteenth of the noise
Dubai, UAE, October 2026 – UAE-founded AHOY Labs (aka Trouvé), part of the AHOY technology group, today announced that its sovereign-by-design AI system has outperformed the enterprise AI search products of OpenAI, Amazon, Microsoft, Google and NVIDIA on EnterpriseRAG-Bench, a public, open-source benchmark for enterprise AI.
The regional deep-tech company surpassed all five across every measure reported by the benchmark and ranked eighth worldwide among 26 systems.
The benchmark gives each AI system the full archive of a simulated company including 500,000 emails, chat threads, files, tickets, sales records and meeting transcripts. It then asks 500 questions. Some answers are spread across several documents. Some documents contradict each other. Some questions have no answer at all. Systems are scored on whether their answers are correct and complete and on whether they bring back only the documents that matter.
“We built AHOY to prove that sovereignty and world-class performance are not a trade-off. Outperforming the biggest names in technology, on both accuracy and precision, shows that regional organisations can keep full control of their data without settling for second-best AI. This is a milestone, not a finish line. Our goal is the top spot,” said Jamil Shinawi, Founder and Group CEO of AHOY.
A fifteenth of the noise:
The biggest gap is in precision. For each answer, OpenAI File Search returned an average of 15.7 irrelevant documents. AHOY Labs returned 1.03, about a fifteenth as many and the fewest of all six systems. Irrelevant documents are noise: they bury the real evidence, add cost and slow people down. They also make it harder to check where an answer came from. With AHOY, the evidence behind an answer is the evidence that matters. For governments, regulators and organisations that must defend every decision they make with AI, that difference is decisive.
Built for sovereignty
AHOY’s platform is designed to run inside the customer’s own premises, on the customer’s own hardware. It can operate fully offline, so sensitive data never leaves the organisation, and AHOY’s internal testing shows it delivers the same benchmark result when running fully offline. Sovereignty is built into the system itself, not added as a hosting option, so every AHOY deployment is sovereign. This makes AHOY Labs the highest-ranked sovereign system on the leaderboard and suited to sectors where data cannot go to a public cloud, such as government, defence, healthcare and energy.
EnterpriseRAG-Bench was created by Onyx and published with an accompanying research paper. It is open source under the MIT licence. Onyx reviews results before posting them to the public leaderboard and does not rank its own product.
Overall Score of Systems:
| System | Overall score | Questions answered correctly | Irrelevant documents per answer (lower is better) |
| AHOY Labs (aka Trouvé) | 74.95 | 79.6% | 1.03 |
| OpenAI File Search | 61.03 | 69.8% | 15.7 |
| Amazon Q (Kendra) | 48.96 | 55.4% | 1.49 |
| Microsoft Azure AI Search | 48.42 | 56.4% | 3.25 |
| Google Vertex AI Search | 41.87 | 49.2% | 4.05 |
| NVIDIA AI Blueprints | 37.73 | 59.6% | 7.72 |
About AHOY:
AHOY is a deep-tech frontier lab and infrastructure company delivering Physical AI: infrastructure that perceives, decides, and operates the physical world. The company’s modular stack runs on-chip, at the edge, on-prem, and in hybrid cloud configurations to provide auditable action for cities, transportation networks, utilities, sovereign sites, and other complex operational environments. Founded by Jamil Shinawi in 2018, AHOY builds the systems that move, sense, and secure the physical world, operating from offices in Canada, Saudi Arabia, the UAE, the U.K. and others.
About AHOY Machine Studio (AMS)
AHOY Machine Studio (AMS) is AHOY’s unified platform for building, governing and deploying Real-World AI. AMS Core is the control plane for identity, policy, signed releases and audit. AMS Forge is where teams and partners build, test and approve models, agents and applications. AMS Anvil deploys and operates them in the field. AMS is chip-agnostic and runs anywhere from cloud-connected setups to fully offline, air-gapped installations on the customer’s own infrastructure, so organisations keep full ownership of their data, models and IP.
Rankings and scores are as published on the EnterpriseRAG-Bench leaderboard on 3 October 2026 and may change. Third-party product names are trademarks of their respective owners. No endorsement is implied.
