What one run costs
A typical run of this task sends about 1.2K tokens and gets back 400, over a single call — 1.6K tokens in total. Those are the numbers below, against the published prices.
| Model | Vendor | Per run | Price evidence |
|---|---|---|---|
| GPT-5 Nano | OpenAI | $0.00022 | Vendor page |
| Gemini 2.5 Flash Lite | $0.00028 | Vendor page | |
| GPT-4.1 nano | OpenAI | $0.00028 | Corroborated |
| Claude Haiku 5.5 | Anthropic | $0.00032 | Vendor page |
| GPT-6 Luna | OpenAI | $0.00032 | Vendor page |
| Jamba Mini | AI21 | $0.00040 | Vendor page |
| GPT-4o mini | OpenAI | $0.00042 | Corroborated |
| GPT-5.6 Luna | OpenAI | $0.00072 | Vendor page |
The cheapest task on this site, and the one where people most often overpay. A reply is a thousand tokens in and a few hundred out, with no multi-step reasoning anywhere in it.
The thing that improves output here is not a bigger model. It is two or three examples of emails you have actually sent, pasted into the prompt, so it writes like you rather than like a press release.
Watch the output rate rather than the input rate: on a task this short, what you are charged for is almost entirely the reply.
What we would pick
This section is our judgement, not a figure read off a page. Everything above is arithmetic on published prices; this is an opinion, and it is labelled as one.