Referencia
Guías
Cómo se hacen los números de overfeed.news: qué significa un precio de alquiler de GPU, cómo se calcula el punto de equilibrio entre API y GPU, cómo leer los precios de API de LLM y cómo se miden los datos.
How GPU rental prices work, and how we measure them
What a price per GPU-hour actually means on RunPod and Vast.ai, why the cheapest listing is not the market price, and why we order cards by VRAM instead of by price.
En esta guía
- Two kinds of price
- Why the cheapest listing is not the price
- Per GPU, not per machine
- Ordered by memory, not by price
- 0,027 US$
- 91
API or your own GPU: the break-even, explained
How we compute the tokens-per-hour above which renting a card beats paying per token for an open-weight model, what goes into the number, and what it deliberately leaves out.
En esta guía
- The question
- The arithmetic
- What the number does not say
- Why prices change the answer
- 459
- 0,027 US$
LLM API prices, explained: per-million tokens, context windows and free tiers
How to read a price per million tokens, why input and output are charged differently, what a context window costs you, and what a free model on OpenRouter does and does not promise.
En esta guía
- Per million tokens
- What a token is
- Context windows
- Free models
- 459
- 16
How overfeed.news collects and publishes its data
The sources we poll, how often, what we keep of each item and why only an excerpt, and how prices are snapshotted.
En esta guía
- Sources and cadence
- Why only an excerpt
- Price snapshots
- Corrections
- 26
- 15.997