Blackbox AI Achieves #1 Ultra Inference Speed and Releases Nemotron 3.5 Lightning
Blackbox AI was verified by Artificial Analysis as the #1 provider for Nemotron 3 Ultra, achieving 454 tokens/sec with zero data retention and enterprise PII stripping. The company also announced Nemotron 3.5 Lightning delivering reasoning output at 1,200 tokens/sec.
Verified State Diff
Impact & Verification Analysis
Enterprise software developers, machine learning engineers, and security compliance teams deploying production AI agents and LLMs.
Delivers industry-leading inference throughput at a 2.7x cost reduction compared to alternative cloud providers, meeting strict security requirements through complete data isolation and zero retention guarantees.
Full Fact Overview
Blackbox AI Technologies Inc. expanded its enterprise inference capabilities with dedicated single-tenant isolated deployments and updated the Blackbox Router to access over 300 open and closed models via a unified endpoint. Independent testing by Artificial Analysis confirmed Blackbox as the fastest provider for NVIDIA Nemotron 3 Ultra at 454 tokens/second—outperforming Nebius (351 t/s) by 30% and CoreWeave (222 t/s) while offering 2.7x lower cost. The enterprise platform also introduced automatic PII stripping prior to prompt delivery to closed models, enforced zero data retention (ZDR), end-to-end encryption, and previewed Nemotron 3.5 Lightning at 1,200 tokens/sec.