# LLM Pricing

> Compare published LLM API rates per million tokens and price a scenario at any cache-hit rate. The checked date and each vendor source sit on the table.

- URL: https://bytescope.dev/en/llm-pricing/
- Locale: en · Also: [ja](https://bytescope.dev/ja/llm-pricing.md) · [zh-Hant](https://bytescope.dev/zh-Hant/llm-pricing.md)
- Runs: The rate card is read in this tab. Nothing is uploaded and no model API is called.
- Requires: Any modern browser (Chrome, Edge, Firefox, Safari), desktop and mobile.
- Updated: 2026-09-08

## What it does

The table is the checked rate card: input, output, cache read, a five-minute cache write, a one-hour cache write, the context window, max output when the card publishes one, the batch factor when it publishes one, and a mark when a long-context tier exists. An empty cell is a line the vendor did not publish. It is not a zero, and it is not a guess. Every number in the table is read from that card.

The calculator is the part a price list does not do. Pick one of the four scenarios, move the cache-hit slider, and tick the models you want side by side. One model from each vendor starts ticked. The totals come from the same cost function the model pages use. Batch is shown in the table and is not applied to the totals, because a batch job is a different purchase from the interactive loop.

Every rate here was confirmed on 2026-09-22, and the link on each row is the vendor page that confirmation used. A model that quotes a price without a date is how stale numbers spread. This page keeps that date with the numbers wherever they go, including the structured data and the Markdown copy.

## Limits

Does not call a model API, does not read agent logs, and does not apply the batch factor to the scenario totals. Rates are the checked card, not a live fetch.

## FAQ

### Why is the date above the table?

Prices on the open web go stale, and a model will repeat an old number without noticing. The date is when this card was last checked against each vendor's own page. The link on the row is that page.

### Does the scenario include the batch discount?

No. Batch is a different way to buy the same tokens, shown as a factor when the card has one. The calculator prices the interactive loop. Folding the factor into the total would make the two purchases look like one.

### What is the cache hit rate?

The share of the repeated context that is read from cache instead of written again. Zero means every turn writes the prefix. Ninety means most of it is a read. A card with no write price still accounts for the miss, billed as input, and says so.