Does an OpenAI API key cost money?
The API key itself does not have a separate purchase price. Charges come from the API usage attached to the account, including applicable model token rates and any separately priced provider features.
LLM API
OpenAI API Pricing and Token Costs
Creating an OpenAI API key does not itself add a fee. API usage is billed separately by model and can include input, cached input, cache writes, and output tokens.
Decision summary: Choose OpenAI when ecosystem maturity and multimodal coverage matter; control cost with routing and output limits.
OpenAI API pricing per token
OpenAI publishes text rates per 1 million tokens. The table below also converts each tracked rate to 1,000 tokens so small requests are easier to estimate.
Cheapest OpenAI API model
GPT-4o mini has the lowest tracked default input and output rates in this table: $0.15 input and $0.60 output per 1M tokens. Test task quality before routing production traffic.
Pricing overview
Usage-based API pricing by model, input tokens, output tokens, context tier, caching, and modality.
Source-tracked model data
OpenAI API model prices in one table
Example cost uses 1,000,000 input tokens and 300,000 output tokens at each model's default tracked rate, with no cache discount.
| Model | Input price | Cached input | Output price | Context | Example cost |
|---|---|---|---|---|---|
| GPT-5.6 Luna Stable | $0.20 / 1M $0.0002 / 1K | $0.02 / 1M $0.00002 / 1K | $1.20 / 1M $0.0012 / 1K | 1.1M tokens | $0.56 |
| GPT-5.6 Sol Stable | $5.00 / 1M $0.005 / 1K | $0.50 / 1M $0.0005 / 1K | $30.00 / 1M $0.03 / 1K | 1.1M tokens | $14.00 |
| GPT-5.6 Terra Stable | $2.00 / 1M $0.002 / 1K | $0.20 / 1M $0.0002 / 1K | $12.00 / 1M $0.012 / 1K | 1.1M tokens | $5.60 |
| gpt-5.4-mini Stable | $0.75 / 1M $0.00075 / 1K | $0.075 / 1M $0.000075 / 1K | $4.50 / 1M $0.0045 / 1K | 400K tokens | $2.10 |
| gpt-5.4-nano Stable | $0.20 / 1M $0.0002 / 1K | $0.02 / 1M $0.00002 / 1K | $1.25 / 1M $0.00125 / 1K | 400K tokens | $0.575 |
| GPT-4o Stable | $2.50 / 1M $0.0025 / 1K | $1.25 / 1M $0.00125 / 1K | $10.00 / 1M $0.01 / 1K | 128K tokens | $5.50 |
| GPT-4o mini Stable | $0.15 / 1M $0.00015 / 1K | $0.075 / 1M $0.000075 / 1K | $0.60 / 1M $0.0006 / 1K | 128K tokens | $0.33 |
Cost boundary: this example compares token charges only. It does not assume equal output quality, latency, retry rates, or output length across models.
Monthly cost calculator
Estimate monthly OpenAI API cost
Choose a model and enter a typical request. The estimate uses the same source-tracked pricing records as the table above.
Estimated monthly token cost$0.00
- Input
- $0.00
- Output
- $0.00
- Per 1,000 requests
- $0.00
Estimate excludes untracked provider-specific charges.
Compare with other providersOfficial price history
OpenAI API recorded model price changes
Verified changes for exact models and pricing profiles available through this provider. New model generations are not treated as price changes.
OpenAI · Standard · Short Context
GPT-5.6 Luna
| Rate | Earlier price | Later recorded price | Change | Percentage |
|---|---|---|---|---|
| Input / 1M | $1.00 | $0.20 | -$0.80 | -80.0% |
| Cached input / 1M | $0.10 | $0.02 | -$0.08 | -80.0% |
| Output / 1M | $6.00 | $1.20 | -$4.80 | -80.0% |
| Cache write / 1M | $1.25 | $0.25 | -$1.00 | -80.0% |
StackLens first observed the lower price on the current official OpenAI pricing source; the provider's original effective date was not independently established.
OpenAI · Standard · Long Context
GPT-5.6 Luna
| Rate | Earlier price | Later recorded price | Change | Percentage |
|---|---|---|---|---|
| Input / 1M | $2.00 | $0.40 | -$1.60 | -80.0% |
| Cached input / 1M | $0.20 | $0.04 | -$0.16 | -80.0% |
| Output / 1M | $9.00 | $1.80 | -$7.20 | -80.0% |
| Cache write / 1M | $2.50 | $0.50 | -$2.00 | -80.0% |
Long-context rates are derived from OpenAI's documented 2x input and 1.5x output multipliers applied to the current base rates.
OpenAI · Standard · Short Context
GPT-5.6 Terra
| Rate | Earlier price | Later recorded price | Change | Percentage |
|---|---|---|---|---|
| Input / 1M | $2.50 | $2.00 | -$0.50 | -20.0% |
| Cached input / 1M | $0.25 | $0.20 | -$0.05 | -20.0% |
| Output / 1M | $15.00 | $12.00 | -$3.00 | -20.0% |
| Cache write / 1M | $3.13 | $2.50 | -$0.625 | -20.0% |
StackLens first observed the lower price on the current official OpenAI pricing source; the provider's original effective date was not independently established.
OpenAI · Standard · Long Context
GPT-5.6 Terra
| Rate | Earlier price | Later recorded price | Change | Percentage |
|---|---|---|---|---|
| Input / 1M | $5.00 | $4.00 | -$1.00 | -20.0% |
| Cached input / 1M | $0.50 | $0.40 | -$0.10 | -20.0% |
| Output / 1M | $22.50 | $18.00 | -$4.50 | -20.0% |
| Cache write / 1M | $6.25 | $5.00 | -$1.25 | -20.0% |
StackLens first observed the lower price on the current official OpenAI pricing source; the provider's original effective date was not independently established.
StackLens assessment
How to shortlist an OpenAI API model
Begin with the lowest-cost tracked route that can pass the workload's acceptance test, then escalate only the cases that need another model.
Start with lower-cost routes
GPT-4o mini and gpt-5.4-nano
Use a representative test set to determine whether the lower tracked token rates are sufficient for routine requests.
Test middle price tiers
gpt-5.4-mini and gpt-5.6-luna
Compare these routes on the lower-cost model's failure cases, including retries and generated output length.
Escalate selected workloads
GPT-4o, gpt-5.6-terra, and gpt-5.6-sol
Reserve higher tracked rates for workloads where measured task outcomes justify the additional cost.
What affects cost
- Output length
- Repeated context
- Long-context tier selection
- Cache write behavior
- Retries and agent loops
- Regional processing premiums
Lower-cost options from the same provider
- gpt-5.4-nano
- gpt-5.4-mini
- gpt-5.6-luna
- GPT-4o mini
Alternative providers or products
- Claude Haiku 4.5
- Gemini 2.5 Flash-Lite
- DeepSeek V4 Flash after workflow testing
Best for
- production AI features
- multimodal apps
- teams that value ecosystem depth
Not ideal for
- teams optimizing only for lowest cost
- workloads without token monitoring
- buyers that need fixed monthly spend
StackLens assessment
How to reduce monthly OpenAI API cost
A useful OpenAI budget starts with request routing, output limits, repeated context, and retry behavior rather than one blended token estimate.
Route simple tasks to lower-cost models
Route simple tasks to lower-cost tracked models such as gpt-5.4-nano, gpt-5.4-mini, gpt-5.6-luna, or GPT-4o mini where quality is good enough.
Cap output tokens
Output tokens can dominate spend because generated answers are billed separately from input.
Cache repeated context
Repeated system prompts, policy text, and document context can inflate input spend when sent on every request.
Monitor retries and agent loops
Retries, tool calls, and agent loops can multiply cost even when each individual request looks small.
Measure cost per feature
Track cost by product feature or workflow so expensive paths are visible before the monthly bill arrives.
Compare before and after routing
Run the same workload through the original route and a lower-cost route, then compare quality and spend together.
Output length can dominate total cost
Example using GPT-4o mini tracked pricing with 1,000,000 input tokens. The only change is output volume.
300,000 output tokens: $0.15 input + $0.18 output = $0.33
2,000,000 output tokens: $0.15 input + $1.20 output = $1.35
Common budget mistakes to avoid
- Choosing a flagship model for every request before measuring quality needs.
- Letting output length grow without product-level caps.
- Sending the same long context repeatedly instead of caching or trimming it.
- Ignoring retries and agent loops in monthly forecasts.
FAQ
OpenAI API pricing questions
Is OpenAI the cheapest LLM API?
Not for every workload. GPT-4o mini and gpt-5.4-nano have low tracked token rates, while larger OpenAI models cost more. Compare the models that pass your task test using the same input volume, output volume, and retry assumptions.
Which OpenAI cost drivers should I measure first?
Measure generated output, repeated context, retries, and agent loops. Then separate routine requests that can use a lower-cost route from the smaller set that needs a higher-priced model.
Does an OpenAI API key have a separate cost?
StackLens does not add a separate API-key fee to its estimates. The calculator models tracked token usage; account, billing, and any non-token charges should be checked against current official OpenAI terms.
How do I estimate OpenAI API cost per request?
Multiply input tokens by the model input rate, multiply output tokens by the output rate, then divide each per-million rate by 1,000,000. Add the two results and account for caching, retries, and other tracked charges.
Is ChatGPT subscription pricing the same as OpenAI API pricing?
No. This page covers API token usage, not ChatGPT subscription pricing. Treat ChatGPT plans and API billing as separate purchasing decisions and verify current account terms before purchase.
Related pages
calculator / llm api costguides / estimate monthly llm api costcalculator / workloads / classification routingcalculator / workloads / customer support chatbotcalculator / workloads / ai agent loopstools / token countermodelspricing / anthropic apicompare / openai vs anthropicalternatives / openai api