News for hackers. Stuff that matters.
Re:Show HN: An unmetered LLM API–$6/month, no token tracking... by yolo-auto on Monday July 06, 2026 @10:08PM (#48812895)
1x MI300x , rocm/sglang , highly optimized image, FP8, FP16 Cache (this fixes looping issues with qwen3.6-35b

~100tok/s average

There are no comments on this comment yet.