3 ms·
From submitter: > here's GPT-OSS-20b running 100 agents simultaneously at 131,076 token context, each agent with its own 131k context window, each hitting sub-
by diggan 1y ago
From submitter:
> here's GPT-OSS-20b running 100 agents simultaneously at 131,076 token context, each agent with its own 131k context window, each hitting sub-100ms ttft, blowing everything out of the sky at 10k tokens/second.