myrtle.ai's VOLLO Sets New STAC-ML Records for Gradient-Boosted Tree Inference
At the STAC Summit in London, myrtle.ai announced that its VOLLO inference accelerator has set new records on the STAC-ML Markets (Inference) benchmarks. The system achieved sub-2µs p99 latency and up to 50 million inferences per second using an AMD Alveo V80LL Compute Accelerator within a Blackcore server. These results represent over a 30% reduction in 99th-percentile latency and at least five times higher throughput compared to previous benchmarks. The performance improvements enable trading firms to use larger, more accurate models without compromising speed or missing market opportunities. Peter Baldwin, CEO of myrtle.ai, emphasized that developers can now test their own models on VOLLO with no FPGA expertise required. VOLLO has already been deployed in live trading environments and is used by leading financial institutions globally. The benchmark results were independently audited by STAC and are available via the STAC Report (SUT ID MRTL2026905).
Tickers: $AMD
Topics: AI inference benchmarking financial trading technology low latency computing