Skip to content
llmwaves

Contact

One mailbox, read by a person

No form, no ticket queue, no chatbot. Everything below arrives in the same place — the prefilled subject lines just make it sortable at the other end.

Email contact@llmwaves.com

Corrections are answered first, usually within a few working days.

What to write about

Six things worth an email

Anything not on this list is welcome too — these are just the ones that come up often enough to have a subject line.

A number looks wrong

The most useful mail we get. Prices change without announcement and benchmark results get revised, so corrections are read first and fixed ahead of everything else.

Include

  • The page it is on, and which figure
  • What it should say instead
  • A link to the source, and the date you saw it

contact@llmwaves.com · “Correction

Press and analysts

Quoting a chart or a figure needs no permission — attribution and a link is enough. Write if you need context on how something was measured, or a version of a chart at a size that reproduces well.

contact@llmwaves.com · “Press

Pitching an article

The blog takes outside contributions. Read the guidelines first: they set out what gets published, what does not, and how a pitch should look.

Read the contributor guidelines

contact@llmwaves.com · “Pitch

Data and licensing

For research use of the index in bulk, or to reproduce a large part of it somewhere else, ask here rather than pointing a scraper at the site. Requests that come with a described use case are usually straightforward.

contact@llmwaves.com · “Data request

Privacy

Questions about what is collected, or a request to see or delete what is held about you. The honest answer is usually that there is very little to look up.

contact@llmwaves.com · “Privacy request

A model or provider is missing

New models are added continuously, but the index is not exhaustive. Tell us what is missing and where its published benchmarks and pricing live.

contact@llmwaves.com · “Coverage

Before you write about a figure

Two things account for most of the mail that turns out not to be an error. The methodology page explains both: cost per task is an estimate computed from list prices against one fixed token profile, and output speed is a median across every provider serving a model rather than the best endpoint available.

An em dash in a table means a figure is missing, not zero. Missing values sink to the bottom of every ranking rather than being counted as nothing.

What we cannot help with

  • Support for the models themselves — billing, API keys, rate limits and outages belong to the vendor, not to an index of them.
  • Guest post placements, link exchanges and paid inserts. The blog publishes contributed writing, but nothing on this site is for sale.
  • Bespoke benchmarking of a private model or a private workload.