Signal65
Third-party observer who verified and validated the 1.1M tokens/second inference record and provided performance analysis
How media typically covers Signal65
Directly quoted in these articles
Microsoft Azure ND GB300 v6 virtual machines achieve a record 1.1 million tokens per second on Llama2 70B inference, a 27% improvement over the previous ND GB200 v6 record and 5× higher throughput per GPU than previous-generation H100 systems.
“Third-party observer who verified and validated the 1.1M tokens/second inference record and provided performance analysis”