USD per 1,000,000 tokens · retrieved 2026-09-27
These are list prices copied from the vendor pages linked below. A dash means that cell was not on the page we copied. Looptally does not fill a dash with “half of standard” or “the usual 0.1×.”
| Model | Input | Cache read | 5m write | 1h write | Output | Storage / hour | Batch input | Batch output |
|---|
OpenAI’s batch cached-input cell is its own number when the table prints one. GPT-5.4 batch cached input is published as $0.13, which is not the exact half of $0.25. Gemini batch cache reads are not uniformly half: Gemini 3.8 Flash lists a halved cache read, and Gemini 2.5 Flash lists the same $0.03 cache read on the batch tier.
Anthropic’s batch table lists input and output. The pricing page says the batch discount stacks with cache multipliers, so those cache cells are the published cache prices times one half, and the model note says so.