Published token rates
Read a dated snapshot of input and output prices for models on OpenRouter, Fireworks, and Together. Credits, keys, and requests stay on the provider you pay. This page does not sell a balance or share one across providers.
Rates were read on 2026-10-07. Confirm the live price on the provider site before you pay.
USD per 1 million tokens, read on 2026-10-07.
Showing models.
Only Together rows in this snapshot have a function-calling mark.
A shared model name is not the same endpoint.
OpenRouter, Fireworks, and Together rows were copied from public pages on 2026-10-07.
These models did not have one price across two Together pages.
Each balance pays only that provider’s own products. This page cannot show your balance, and it does not list a discount.
No balance is stored here.
Fee is charged when you buy credits. Model rates are passed through.
OpenRouter billing rules.
No balance is stored here.
Prepaid credits also pay for deployments and training.
Minimum, fee, and refund were not printed on the source page.
Fireworks billing rules.
No balance is stored here.
First credit purchase is $15. Later purchases can be $5.
Together billing rules.
Compare the copied input, output, cache, and context. Then open the source page and check the live number.
Pay on that provider’s account. A credit on OpenRouter does not call Fireworks or Together, and the reverse is also true.
The key is shown by the provider. Aurelian never asks for it and cannot store it.
Use the endpoint and model id from the provider’s current guide. This site does not forward the request.
Chat, coding, and writing on these pages use the text token rates. Image generation is a different unit.
Billed as input and output tokens on the text rows.
No separate coding price was printed on the pages we read.
Drafts and summaries use the same input and output token rates as chat.
Together’s docs bill some image models per image and others per megapixel.
ImagesCopied from the Together docs catalog on 2026-10-07.
Create the key on the provider, then send the model id that provider documents today.
OpenRouter posts to this URL.
Endpointhttps://openrouter.ai/api/v1/chat/completions
Fireworks uses this inference URL.
Endpointhttps://api.fireworks.ai/inference/v1/chat/completions
Together posts here.
Endpointhttps://api.together.ai/v1/chat/completions
Each provider deducts usage from credits you buy there.
A dated subset, not a full catalog.
This page cannot refund anything.
Call the provider endpoint with a key created there.
A saved language wins. Otherwise a country lookup runs, and English is the fallback.
No.