Services / Catalogue feed diagnostic / Cost calculator
What does AI enrichment cost per 1,000 listings?
Enter your catalogue size, how many images and how many words each listing gets, and pick a model. The estimate uses each developer's own list price.
GPT-6.1 Terra
$3.81 per 1,000 listings
$19.06 for 5,000 listings, with the batch discount.
OpenAI's list price: $1.00 per million input tokens, $5.00 per million output, checked 2026-11-05 (source).
Cost is the smaller question. A model that invents a material or a size costs more in returns, rejected listings and review time than any difference on this page. Check accuracy on your own products before you pick on price.
Ask for a free 20-product spot checkCost per 1,000 listings, by model
| Model | Developer | Per listing | Per 1,000 listings | Total |
|---|---|---|---|---|
| Gemini 2.5 Flash Lite | $0.00036 | $0.36 | $1.80 | |
| GPT-6 Luna | OpenAI | $0.00038 | $0.38 | $1.91 |
| MiMo-V2.6-Flash | Xiaomi | $0.00045 | $0.45 | $2.24 |
| MiniMax M3.1 Flash | MiniMax | $0.00054 | $0.54 | $2.71 |
| GPT-5.6 Luna | OpenAI | $0.00080 | $0.80 | $4.01 |
| GPT-5.4 Nano | OpenAI | $0.00081 | $0.81 | $4.07 |
| DeepSeek V4.1 Flash | DeepSeek | $0.0011 | $1.08 | $5.42 |
| DeepSeek-V4.1-Flash-1017 | DeepSeek | $0.0011 | $1.08 | $5.42 |
| MiniMax M3 | MiniMax | $0.0011 | $1.08 | $5.42 |
| Gemini 2.5 Flash | $0.0013 | $1.35 | $6.73 | |
| Gemini 3.5 Flash Lite | $0.0013 | $1.35 | $6.73 | |
| MiMo-V2.6-Pro | Xiaomi | $0.0014 | $1.39 | $6.97 |
| GPT-5.4 Mini | OpenAI | $0.0030 | $3.01 | $15.06 |
| Kimi K2.6 | Moonshot AI | $0.0035 | $3.47 | $17.35 |
| Kimi K2.7 Code | Moonshot AI | $0.0035 | $3.47 | $17.35 |
| Claude Haiku 4.5 | Anthropic | $0.0038 | $3.81 | $19.06 |
| GPT-6.1 Terra | OpenAI | $0.0038 | $3.81 | $19.06 |
| Muse Spark 1.3 | Meta | $0.0044 | $4.36 | $21.80 |
| Gemini 2.5 Pro | $0.0055 | $5.53 | $27.63 | |
| Gemini 3.6 Flash | $0.0057 | $5.72 | $28.59 | |
| Gemini 3.7 Flash | $0.0057 | $5.72 | $28.59 | |
| Gemini 3.8 Flash | $0.0057 | $5.72 | $28.59 | |
| Gemini 3.5 Flash | $0.0060 | $6.02 | $30.11 | |
| Claude Sonnet 5 | Anthropic | $0.0076 | $7.62 | $38.12 |
| Claude Sonnet 5.5 | Anthropic | $0.0076 | $7.62 | $38.12 |
| GPT-6 Sol | OpenAI | $0.0076 | $7.62 | $38.12 |
| GPT-6.1 Sol | OpenAI | $0.0076 | $7.62 | $38.12 |
| Gemini 3.1 Pro Preview | $0.0080 | $8.03 | $40.15 | |
| GPT-5.6 Terra | OpenAI | $0.0080 | $8.03 | $40.15 |
| GPT-5.4 | OpenAI | $0.01 | $10.04 | $50.19 |
| Kimi K3 | Moonshot AI | $0.01 | $11.44 | $57.19 |
| Claude Opus 5.5 | Anthropic | $0.02 | $15.25 | $76.25 |
| Muse Wonder | Meta | $0.02 | $15.25 | $76.25 |
| GPT-5.6 Sol | OpenAI | $0.02 | $15.25 | $76.25 |
| Claude Opus 5 | Anthropic | $0.02 | $19.06 | $95.31 |
| GPT-5.5 | OpenAI | $0.02 | $20.07 | $100 |
| Claude Fable 5 | Anthropic | $0.04 | $38.12 | $191 |
| Claude Fable 5.1 | Anthropic | $0.04 | $38.12 | $191 |
| GPT-6 Astra | OpenAI | $0.04 | $38.12 | $191 |
Models that read images and whose developer list price we have read, cheapest first at your inputs. Prices are each developer's own list price, read by hand from its pricing page; see LLM API pricing for every model and date checked. They don't include taxes, promotions or volume deals.
How we work it out
- Input per listing: about 800 tokens for the instructions and the product's feed row, plus about 1,000 tokens per image.
- Output per listing: about 1.35 tokens per English word written.
- Cost: input tokens at the input price plus output tokens at the output price, less the batch discount if you ticked it and the developer offers one.
What it leaves out
- Thinking. Reasoning tokens bill as output. With thinking on, a listing can cost several times more.
- Image size. Developers count image tokens differently, and larger images cost more. Check your provider's documentation for your image size.
- Retries and review. Failed calls, second passes and the people who check the output aren't counted.
- Prompt caching. Repeated instructions can cost less with caching. We don't assume it.
Find out which setup you can trust, not only what it costs.
Send us 20 of your products and your current AI output. We score them with the CatalogBench checks and send back a one-page result, free.