OpenAI API Pricing Calculator
Compare GPT-5.5, Claude, Gemini, DeepSeek, Llama and Cohere API prices side by side — then calculate the cost for your own monthly token volume.
Prices checked manually · 6 Jul 2026
Quick answer: OpenAI API cost = (input tokens ÷ 1,000,000 × input price) + (output tokens ÷ 1,000,000 × output price). As of July 2026, GPT-5.5 costs about $5 per 1M input tokens and $30 per 1M output; GPT-5.4 is $2.50 / $15; the budget baseline GPT-4o-mini is $0.15 / $0.60. Your real monthly cost depends on the model, your input:output mix and request volume — the calculator below computes it across every major provider on a blended per-million-token basis.
Monthly cost = (input tokens ÷ 1,000,000 × input price)
+ (output tokens ÷ 1,000,000 × output price)Example — GPT-4o-mini at 5M input + 2M output tokens: (5 × $0.15) + (2 × $0.60) = $0.75 + $1.20 = $1.95 / month.
| Model | Input /1M | Output /1M | Blended* | Context | × vs 4o-mini▲ |
|---|
Drag the mix · click a column to sort · click a row to load it into the calculator ↓
Cost for your volume
gpt-4o-miniOr start from a typical workload:
Different apps have different token profiles: a summarisation or RAG app is input-heavy (large context, short output), while an AI agent or coding assistant is output-heavy (long generations, multiple calls). Pick a workload above to see how the cheapest model changes for your case.
*Blended reflects the input:output mix set above. **Assumes ~7K tokens/request. The × vs 4o-mini column shows how many times more (or less) expensive a model is than GPT-4o-mini at the chosen mix — the "cheapest" badge still marks the single lowest-cost model. Estimates only — verify against your provider invoice.
How to use this OpenAI API pricing calculator
Pick any model in the comparison table to load it into the calculator, then enter your monthly input and output token volume. The tool shows the estimated monthly, daily and yearly cost, plus a blended per-million-token rate. The comparison table ranks every major model by how its blended price compares to GPT-4o-mini — the popular budget baseline — so you can see at a glance whether a model is cheaper than 4o-mini or a 40× premium on it.
Why the input:output mix changes the answer
Most providers charge more for output tokens than input tokens — often 3× to 5× more. That means the "cheapest" model depends on how your workload is split. A summarisation app that reads a lot and writes a little has a very different blended cost from an agent that generates long responses. Drag the input:output handle above (base is 3:1) to match your own workload, and both the blended column and the × vs 4o-mini ranking recompute instantly.
Input vs output token pricing, explained
Input tokens are what you send to the model — your prompt, context and any retrieved documents. Output tokens are what the model generates back. Because generation is more compute-intensive, output is priced higher. When you estimate cost, you have to weight the two separately; a single blended number only works once you fix a mix ratio, which is exactly what the handle above lets you set.
Fractional CFO for AI SaaS founders. These API prices are the raw input to the gross-margin and pricing work I do with founders preparing for their first round. More about Dmitry →
Frequently asked questions
OpenAI API cost depends on the model and your token volume. As of 6 July 2026, the flagship GPT-5.5 runs about $5 per 1M input tokens and $30 per 1M output, and GPT-5.4 about $2.50 / $15, while a small model like GPT-4o-mini is roughly $0.15 input and $0.60 output. A typical app using 5M input and 2M output tokens a month pays a few dollars on a mini model and tens of dollars on a flagship. Use the calculator above for your exact numbers.
Per request, no — but cost scales with token volume and model choice. A flagship model can cost 40× or more than a small model like GPT-4o-mini for the same tokens, so the model you pick matters more than the provider you pick.
The cheapest LLM APIs per token are the smallest models — Cohere Command R7B, DeepSeek V4 Flash, Gemini Flash and Llama — often well under $0.30 per 1M blended tokens. Sort the comparison table by "× vs 4o-mini" to rank models against the popular GPT-4o-mini baseline; the "cheapest" badge always marks the single lowest-cost model.
Multiply your monthly input tokens by the model's input price and your output tokens by its output price, then add them. The calculator above does this automatically and also shows per-day, per-year and per-1,000-request figures.
Prices tell you the cost. They don't tell you the margin.
These numbers are what the API charges. They don't tell you what happens to your gross margin once a small share of heavy users starts burning most of your tokens — or how to model that across 36 months for a VC meeting. That is the AI SaaS Financial Model bundle, built by the same fractional CFO behind this calculator.
See the full AI SaaS model →