#USA #Cambridge #Myrtle.ai #VOLLO #STAC-ML

Myrtle.ai's VOLLO Sets New Records in STAC-ML Inference

On October 6, 2026, at the STAC Summit in London, myrtle.ai announced groundbreaking achievements with its VOLLO® inference accelerator. The device has set new benchmarks in the STAC-ML Markets for Gradient-Boosted Trees, showcasing a remarkable reduction in latency and an impressive increase in throughput.

Key Performance Highlights

The VOLLO accelerator has achieved an extraordinary p99 latency of under 2 microseconds, significantly enhancing the efficiency of financial trading models. Notably, the smallest model reached a throughput of 50 million inference operations per second, with a p99 latency clocking in at just 1.77 microseconds. Such performance metrics are critical in high-frequency trading, where fractions of a second can substantially influence profitability.

In electronic trading, the delay between market data receipt and decision-making directly impacts returns. Consequently, the deterministic low latency provided by VOLLO allows companies to implement larger and more precise models without lagging behind the market. This eliminates the need to sacrifice model quality for speed.

Improving Financial Decision-Making

Peter Baldwin, CEO of myrtle.ai, emphasized the breakthrough, stating, "Trading firms always seek to use more powerful models without sacrificing speed, and these results show that it is possible. Developers can now test their own models on VOLLO without requiring FPGA expertise, thereby experiencing the benefits firsthand."

With the latest STAC Tacana announced results in April, VOLLO has now set records for deterministic latency across decision trees and neural networks alike. Its practical application has already demonstrated real-world efficacy, generating alpha for numerous leading trading firms worldwide during hundreds of thousands of hours of live trading.

Furthermore, the flexibility of VOLLO has made it a preferred platform in varied sectors beyond finance, including telecommunications, cybersecurity, and defense.

Importance of STAC-ML Markets

STAC-ML Markets (Inference) serves as the technological standard for conducting inference computations based on real-time market data. Developed by quantitative analysts and technologists from top financial firms, it provides performance reports, resource efficiency, and quality evaluations of the technology stacks utilized in model execution. The complete results from VOLLO can be accessed in the STAC report (SUT-ID MRTL2026905) at STAC Research.

Moreover, ML developers can already start testing their models against VOLLO without needing specialized FPGA tools or knowledge. Interested parties can visit myrtle.ai/vollo-trees or contact [email protected] to get started.

About myrtle.ai

Myrtle.ai specializes in AI and ML software, delivering inference accelerators with ultra-low latency for FPGA-based platforms from all leading FPGA vendors. The company's accelerators find application across various sectors, including financial trading, telecommunications, large language models, speech processing, and recommendation systems. Notably, VOLLO, VOLLO Accelerator, and the VOLLO logo are registered trademarks of myrtle.ai, while