GPT-4o mini token cost & pricing
GPT-4o mini was the budget workhorse of the GPT-4 generation at $0.15 per million input and $0.60 per million output. It remains one of the cheapest multimodal-capable models available, which keeps it relevant for image classification and tagging pipelines.
GPT-4o mini specifications
| Vendor | OpenAI |
|---|---|
| Input price | $0.15 / 1M tokens |
| Output price | $0.6 / 1M tokens |
| Cached input price | $0.075 / 1M tokens (50% cheaper) |
| Context window | 128,000 tokens |
| Max output | 16,384 tokens |
| Open weights | No |
| Price rank | 3 of 17 (cheapest first) |
| Best for | budget, multimodal |
What GPT-4o mini costs in practice
Four common workloads, priced at list rates. Monthly figures assume 30 days.
| Workload | Per request | Requests/day | Per month |
|---|---|---|---|
| Chatbot · 1K in / 500 out | $0.0004 | 10,000 | $135.00 |
| RAG Q&A · 8K in / 400 out | $0.0014 | 5,000 | $216.00 |
| Doc summary · 30K in / 1K out | $0.0051 | 1,000 | $153.00 |
| Code agent · 60K in / 3K out | $0.01 | 500 | $162.00 |
How many requests does $100 buy?
Assuming a 1,000-token prompt and a 500-token response per request.
How GPT-4o mini pricing actually works
The 4x input-to-output ratio is friendlier than the 8x on the GPT-5 line, so workloads with short prompts and longer answers lose less by staying here. Cached input is $0.075 per million, a 50% discount.
When to use GPT-4o mini
Good for high-volume image tagging, simple vision checks, lightweight chat and text normalization. For pure text it is dominated by GPT-5 nano on price and GPT-5 mini on quality, so its real niche is cheap multimodal work.
How to cut your GPT-4o mini bill
Downscale images before sending them — image tokens scale with resolution, and most classification tasks do not need full resolution. That single change often halves spend on GPT-4o mini vision workloads.
Other OpenAI models
Cheaper alternatives to GPT-4o mini
Run the numbers on your own prompt
Paste your actual text into the token counter to get an exact count instead of an estimate. Nothing is uploaded.
Open the calculator →