Backboard.io announced it has achieved state-of-the-art performance across both leading AI memory benchmarks, a first of it’s kind acheivement.
OTTAWA, ONTARIO, CANADA, February 11, 2026 /EINPresswire.com/ — Backboard.io announced it has achieved state-of-the-art performance across both leading AI memory benchmarks, a rare milestone that signals a meaningful acceleration in how agentic AI systems can be built and deployed in real-world environments.
As interest in agentic AI grows, many systems continue to struggle with a core limitation: memory. Without reliable long-term memory, agents lose context, fail to coordinate, and break down as systems become more complex. As a result, many agent-based systems work in demos but fail in production.
Backboard addresses this challenge by treating memory as foundational infrastructure rather than an add-on.
In independent testing, Backboard led both the LoCoMo and LongMemEval benchmarks, which measure how well AI systems retain, update, and reason over information across long interactions and multiple sessions. Achieving top performance on both benchmarks is uncommon, as most systems optimize for either short-horizon precision or long-horizon persistence, but not both.
An independent evaluation conducted by NewMath, a Texas-based engineering firm and AWS Small Partner of the Year, measured Backboard’s performance on LongMemEval using the benchmark’s original academic specification. Backboard achieved 93.4% overall accuracy, the highest publicly reported result under consistent methodology. Backboard previously reported 90.1% accuracy on LoCoMo, with results publicly available and reproducible via GitHub.
Importantly, Backboard did not set out to optimize for benchmarks. The LongMemEval evaluation was initiated and run independently, and the LoCoMo benchmark was explored simply to understand where Backboard fit relative to existing research.
“We didn’t build Backboard to chase benchmarks,” said Rob Imbeault, founder of Backboard.io. “We built it to solve what actually breaks when AI systems run for long periods of time, across multiple agents, using different models. The benchmarks just happened to confirm that approach.”
During post-evaluation review, Backboard and the independent evaluator identified cases where Backboard’s responses were more precise and semantically accurate than the benchmark’s expected answer, but were marked incorrect due to labeling ambiguity. As a result, the reported LongMemEval score should be considered a conservative lower bound on performance.
Backboard is designed as a unified AI stack, built to work end-to-end while remaining modular by design. Through a single API, the platform provides persistent long-term memory, native embeddings, retrieval-augmented generation (RAG), shared memory across agents, and access to more than 17,000 large language models. Teams can adopt individual components or use the full stack, without sacrificing interoperability or future flexibility.
By unifying memory, retrieval, and multi-model access around a shared, durable memory layer, Backboard makes agentic systems practical to deploy at scale. Agents running on different models can coordinate, retain context, and evolve over time without fragmentation.
“Agentic AI doesn’t become meaningful because you label something an agent,” said Imbeault. “It becomes meaningful when agents can remember, coordinate, and operate over time. Solving memory is what makes that possible.”
Imbeault previously founded Assent, a platform trusted by Fortune 100 companies to manage complex supply chain and regulatory compliance workflows. That experience shaped Backboard’s focus on durability, correctness, and trust from the outset.
Backboard also announced plans to introduce Switchboard, a forthcoming capability designed to help teams better understand how complex AI system configurations behave under real-world constraints. Additional details will be shared in the coming weeks.
More information on Backboard’s benchmarks and technical approach is available on the company’s website and GitHub repository.
Maya Ellis
Backboard IO
hello@backboard.io
Visit us on social media:
LinkedIn
YouTube
X
Legal Disclaimer:
EIN Presswire provides this news content “as is” without warranty of any kind. We do not accept any responsibility or liability
for the accuracy, content, images, videos, licenses, completeness, legality, or reliability of the information contained in this
article. If you have any complaints or copyright issues related to this article, kindly contact the author above.
