25 October 2026 EN ES
The Serving Desk

The stack under your AI product — models, serving, and what it costs

The Serving Desk

Contact

If you run inference in production and your numbers disagree with ours, that is a story. Send the conditions with them.

What to send

  • A benchmark we should reproduce — with hardware, batch size and model version
  • A cost breakdown — what a real workload actually bills at, per month
  • A correction — headline, link and the figure in question

We read everything and reply to what we can act on, usually within a few working days.

Send us a message

Required — we cannot reply without it.

We use your address only to answer you.