Machine Learning · pricing
OpenAI API pricing
OpenAI API publishes 2 tiers. Below is each one, what it adds over the tier beneath it, and where the ladder stops being worth climbing for most people.
As of 30 August 2026, OpenAI API pricing is $0.15/per-million-tokens across 2 published tiers. Figures are the vendor's list prices as recorded on this page's last check; the vendor's site is the final word.
- Entry price
- $0.15/per-million-tokens
- Model
- Usage-based
- Tiers
- 2
- Free tier
- Not on record
OpenAI API plans, side by side
Every published tier in ascending order of price, with the number of features each one lists. An absence here means the record does not itemise it, not that the tier lacks it.
| Plan | Price | Features listed | Step up from the tier below |
|---|---|---|---|
| GPT-4o mini | $0.15/per-million-input-tokens | 2 | Entry tier |
| GPT-4o | $5/per-million-input-tokens | 2 | +$4.85/per-million-input-tokens, 2 more features |
What each tier adds
The set difference between each tier's feature list and the one beneath it. Where a tier lists nothing new, the vendor has either stopped repeating the inherited list or the tier raises limits rather than adding capability, the record does not distinguish the two, so this page does not either.
GPT-4o mini
$0.15/per-million-input-tokensThe entry tier. It covers fast, affordable.
GPT-4o
$5/per-million-input-tokensOver GPT-4o mini, this tier adds:
- Multimodal
- 128K context
What the product covers
The full OpenAI API feature record, grouped as the catalogue groups it. This is the product as a whole; the record does not map every feature onto a specific tier.
Other
- Text and reasoning models
- Embeddings
- Speech and audio
- Image generation
- Function calling
- Structured outputs
- Batch processing
- Prompt caching
- Fine-tuning
- Usage tiers and rate limits
People bring OpenAI API in for adding summarisation, drafting or classification to an existing product where building a model would take longer than the product's whole roadmap, retrieval-augmented question answering over internal documents, using the embedding and generation models together, extracting structured records from unstructured text, where schema-constrained output removes most of the parsing problem, prototyping a language feature quickly to find out whether it is worth the cost of a self-hosted alternative later. If that is not the job you are buying for, the alternatives to OpenAI API are worth a look before you commit to a tier.
How that compares in Machine Learning
Across the 3 machine learning tools in this directory that publish a starting price, the median entry point is $29/month. OpenAI API starts at $0.15/per-million-tokens, which puts it below the middle of its category.
| Tool | Entry price | Model | Rated | Head to head |
|---|---|---|---|---|
| OpenAI API (this page) | $0.15/per-million-tokens | usage-based | - | |
| ClearML | Free | Open-source self-hosted, with paid hosted and enterprise tiers | - | vs OpenAI API |
| Langwatch | Free, then €29/month | Tiered subscription with usage-based overage charges | - | vs OpenAI API |
| Haystack | Free | Open-source with optional paid enterprise support | - | vs OpenAI API |
| Semantic Kernel | Free | Open source, no pricing | - | vs OpenAI API |
| Fal AI | $1.89/hour | usage-based | - | vs OpenAI API |
| H2O.ai | Free | freemium | - | vs OpenAI API |
Ratings are aggregated from third-party sources and imported with each catalogue entry, they are not reviews hosted on Softwr. How each figure is used is set out on the OpenAI API badges page.
Before you pay for OpenAI API
Three things this page cannot tell you, and all three change the bill. The record carries a price and a billing period but not the unit - per-seat and flat-rate pricing are indistinguishable in this data, so a five-person team may be looking at five times the figure above. It carries no annual rate, so any discount for paying yearly is not something this page can quote. And it carries no trial length.
What it can tell you is the shape of the ladder: 2 tiers between $0.15/per-million-input-tokens and $5/per-million-input-tokens, with the jump itemised above. With no free tier on record, the cheapest way to answer the rest is to compare OpenAI API against the tools that do have one before committing.
OpenAI API runs on api, and is published by OpenAI of San Francisco, California, USA. The full record is on the OpenAI API review, and the rest of the category is under best machine learning tools.
OpenAI API pricing questions
- How much does OpenAI API cost?
- OpenAI API publishes 2 tiers, from $0.15/per-million-input-tokens for GPT-4o mini up to $5/per-million-input-tokens for GPT-4o. The cheapest paid tier is $0.15/per-million-input-tokens.
- Does OpenAI API have a free plan?
- Not on the record we hold: OpenAI API is listed as usage-based. Whether a free trial is offered is not something this record captures, so check the vendor's site.
- What is the difference between GPT-4o mini and GPT-4o on OpenAI API?
- GPT-4o costs $5/per-million-input-tokens against $0.15/per-million-input-tokens, and adds multimodal, 128k context.
- Is the GPT-4o plan on OpenAI API worth it?
- That depends entirely on whether you need what it adds, which on the published record is multimodal, 128k context. It costs $5/per-million-input-tokens against $0.15/per-million-input-tokens for GPT-4o mini. Softwr has not used either tier, so this page can tell you what changes, not whether it is worth the difference to you.
- Is OpenAI API expensive for a machine learning tool?
- It starts below the middle of its category. Across the 3 machine learning tools in this directory that publish a starting price, the median entry point is $29/month; OpenAI API starts at $0.15/per-million-tokens.
- Which machine learning tools can I use without paying?
- 7 of the 8 machine learning tools listed alongside OpenAI API have a free tier: ClearML, Langwatch, Haystack, Semantic Kernel, H2O.ai.
- What am I actually paying for with OpenAI API?
- The record lists 10 features across 1 area: other. In practice it is brought in for adding summarisation, drafting or classification to an existing product where building a model would take longer than the product's whole roadmap, retrieval-augmented question answering over internal documents, using the embedding and generation models together, extracting structured records from unstructured text, where schema-constrained output removes most of the parsing problem.
- Does OpenAI API charge per user?
- The record carries a price and a billing period for each of its 2 tiers, but not the unit that price is charged in. Per-seat and flat-rate billing look identical in this data, so check the vendor's page before budgeting for a team.
- Are these OpenAI API prices current?
- They are what the catalogue entry holds, refreshed when the entry is. Vendors change pricing without notice and nobody at Softwr re-verifies each tier by hand, so treat this as a structured summary and the vendor's own pricing page as the authority.
- What should I compare OpenAI API against before paying?
- The closest machine learning tools in this directory are ClearML, Langwatch, Haystack, Semantic Kernel. Each has a side-by-side comparison with OpenAI API covering price, platforms and features.