Output price
It is always higher than input — knowing why helps you choose the right model for your product
What is output price
Output price is what you pay for the model's response.
Input price covers the text you send. Output price covers the text you receive back. They are billed separately, and output is almost always more expensive. For most models, output tokens cost 3–5× more per token than input tokens.
Claude 3.5 Sonnet: $3.00/1M input, $15.00/1M output. GPT-4o: $5.00/1M input, $15.00/1M output. The gap is consistent across providers. Verified March 2026.
Why output costs more
Generating text is harder than reading it.
When a model processes your input, it encodes your tokens and builds context. When it generates output, it does that plus makes a prediction for every single token it writes — one at a time, each prediction depending on all the ones before it. More computation. More cost.
The longer the response you ask for, the more that adds up.
Why this matters to you
The split between input and output price shapes which model is right for your product.
If you send long documents but want short answers — a summarisation product, a classifier, a data extractor — your output volume is low. Input price matters more. A model with low input cost and higher output cost works in your favour.
If you send short prompts but generate long responses — a writing assistant, a code generator, a customer support bot with detailed answers — output volume dominates. You want a model where output price is competitive, even if input costs a little more.
The mistake is optimising for input price when your cost is mostly output, or the reverse. Know your ratio first.
How to use this
Estimate your average response length in words. Divide by 0.75 for tokens. Multiply by your monthly call volume. Now compare output cost across three models — not just input price. The cheapest model on input might not be the cheapest model for your actual usage pattern.
Verified March 2026 · Source: Anthropic, OpenAI pricing pages