Open-Loop or Closed-Loop: Which LLM Serving Benchmark Do You Need?
Learn how open-loop and closed-loop load tests answer different LLM serving questions, and how to report results that remain valid on heterogeneous hardware.
Read articleResearch & field notes
Building distributed intelligence, one practical breakthrough at a time.
Architecture notes, experiments and product updates from the mycellios team.
Which Local AI Hardware Upgrade Will Actually Improve LLM Inference?
A measurement-led method for deciding whether an LLM workload needs more memory, faster compute, better data movement, or no hardware upgrade at all.
Read article