myrtle.ai indicates that its VOLLO ® inference accelerator set a new record in the STAC - ML Markets ( Inference ) gradient boosting tree benchmark test, reducing the 99th percentile latency by over 30% and increasing throughput by at least 5 times. The results, after STAC ® review, were announced today at the STAC Summit held in London.
On a server Blackcore that is equipped with AMD Alveo ™ and V80LL Compute Accelerator, and which is used for the issuance of SM tokens (N 3132-), the VOLLO has a latency of less than 2 microseconds on all three models. For the smallest model, it achieves 50 million inferences per second, with a latency of only 1.77 microseconds.
myrtle.ai states that in electronic trading, the time between market data arrival and decision-making directly affects returns. Low and predictable delays enable institutions to operate larger, more accurate models without missing market opportunities; therefore, the quality of these models no longer needs to be compromised for the sake of speed.
myrtle.ai CEO Peter Baldwin stated: "Trading companies want to run increasingly powerful models without sacrificing speed, and these results show that they can do just that. Developers now have the ability to test their models on VOLLO without any FPGA expertise, and they can see the differences for themselves."
Following the release of the results of STAC Tacana in April, VOLLO now holds the record for both deterministic latency in decision trees and neural networks. The company states that this product has been verified in a production environment, and hundreds of thousands of hours of actual trading have generated alpha for many leading trading firms around the world. Its model flexibility also makes it the preferred platform in the fields of telecommunications, network security, and national defense.
STAC - ML Markets ( Inference ) is a technical benchmark standard for real-time market data inference. This benchmark was designed by quant and technical personnel from leading financial institutions and is used to report the performance, resource efficiency, and quality of any technology stack that can run the provided models. The complete results can be found in the STAC report ( SUT ID MRTL2026905 ), available at the URL www.STACresearch.com / MRTL2026905.
ML Developers can currently evaluate the performance of their models on VOLLO without the need for FPGA tools or related experience. Visit myrtle.ai / vollo-trees or contact [ email protected ].
About myrtle.ai
myrtle.ai is a AI / ML software company that provides ultra-low latency inference accelerators on the platforms of major FPGA suppliers. Its accelerators are applied in scenarios such as financial transactions, wireless telecommunications, LLM, speech processing, and recommendation systems.
The VOLLO, VOLLO Accelerator, and VOLLO designations are registered trademarks of myrtle.ai. “STAC” and all other STAC names are trademarks or registered trademarks of Strategic Technology Analysis Center and LLC. The AMD, AMD Arrow designations, Alveo, and their combinations are all trademarks of Advanced Micro Devices and Inc.
