OpenAI API Pricing: GPT Costs vs ChatGPT Plans
Start with how you will use OpenAI: an application calling an API key is billed differently from ChatGPT or Codex using a subscription. Below, compare tracked GPT model variants, then use the cost formula and official sources to check the bill for your own workload. The table is a database snapshot, not a guarantee of endpoint availability or a complete OpenAI rate card.
Current API price snapshot
Up to 12 tracked language models are shown. Channels are selected by lowest input rate normalized to USD, not lowest total cost. Examples use 1M input plus 250K output tokens, without additional cache, batch or context-tier adjustments or tool fees. Processing-mode variants are not interchangeable; no total is shown when output pricing is unknown.
| Model | Lowest-input-rate channel | Input / 1M | Output / 1M | Example cost (USD) | Channels | Selected price verified |
|---|---|---|---|---|---|---|
| GPT 6.1 Sol Pro Batch | OpenRouter | $1.00 | $5.00 | $2.25 | 1 | Oct 14, 2026 |
| GPT 6.1 Sol Pro | OpenRouter | $2.00 | $10.00 | $4.50 | 1 | Oct 14, 2026 |
| GPT 6.1 Sol Batch | OpenRouter | $1.00 | $5.00 | $2.25 | 1 | Oct 14, 2026 |
| GPT 6.1 Sol | OpenAI | $2.00 | $10.00 | $4.50 | 2 | Oct 14, 2026 |
| GPT 6 Luna Pro Batch | OpenRouter | $0.05 | $0.25 | $0.1125 | 1 | Oct 14, 2026 |
| GPT 6 Luna Pro | OpenRouter | $0.10 | $0.50 | $0.225 | 1 | Oct 14, 2026 |
| GPT 6 Luna Batch | OpenRouter | $0.05 | $0.25 | $0.1125 | 1 | Oct 14, 2026 |
| GPT 6 Luna | OpenAI | $0.10 | $0.50 | $0.225 | 2 | Oct 14, 2026 |
| GPT 6 Sol Pro Batch | OpenRouter | $1.00 | $5.00 | $2.25 | 1 | Oct 14, 2026 |
| GPT 6 Sol Pro | OpenRouter | $2.00 | $10.00 | $4.50 | 1 | Oct 14, 2026 |
| GPT 6 Sol Batch | OpenRouter | $1.00 | $5.00 | $2.25 | 1 | Oct 14, 2026 |
| GPT 6 Sol | OpenRouter | $2.00 | $10.00 | $4.50 | 1 | Oct 14, 2026 |
Does ChatGPT Plus include API usage?
Do not treat a ChatGPT subscription as a balance for API-key calls. OpenAI separates API token billing from subscription usage. Codex can use an eligible ChatGPT plan, while Codex authenticated with an API key follows API pricing. Compare subscription entitlements on the OpenAI plans page; do not divide a monthly plan price by an API token rate to infer an included token allowance.
Calculate a monthly API budget
For ordinary uncached text usage, cost = input tokens ÷ 1,000,000 × input rate + billable output tokens ÷ 1,000,000 × output rate. As an illustrative example, 10,000 requests averaging 1,000 input and 250 output tokens consume 10M input and 2.5M output tokens: ten times the workload shown in each table row. Measure full request usage, including repeated conversation history, rather than counting only the latest message. These volumes are examples, not a usage forecast.
Usage example: repeated GPT prompts
Assume 10,000 standard, short-context text requests, each with 1,000 input and 250 billable output tokens. Assume usage reports confirm 8M cache-hit input tokens and 2M ordinary input tokens, on a model without separately billed cache writes. Let I, C and O be its current ordinary-input, cached-input and output rates per million tokens: the illustration costs 2 × I + 8 × C + 2.5 × O. It excludes tools, retries and regional uplifts, and uses no batch discount. If caching misses, move those tokens back to I; if the model bills cache writes, account for them at their own rate. Use the Compare channels entry and each model link to check the same processing mode before choosing a channel.
Caching, batch and long-context pricing
The official rate card distinguishes ordinary input, cached input, cache writes for applicable models, and output. It also lists processing modes and short- versus long-context rates. Choose the row that matches your model and request, not simply the smallest number on the page. A batch variant is not a like-for-like replacement for an interactive endpoint. Our snapshot does not apply extra discounts or context-tier adjustments to the stored price.
Compare the same model across channels
Match the exact model identifier, version, processing mode and region before comparing OpenAI with a cloud or aggregator. The table selects the lowest tracked input rate for each model after currency conversion; it does not rank channels by your total bill. A channel with cheaper input but more expensive output can cost more for generation-heavy work. Open the model detail page to compare the other channels and verify their terms, limits and sources.
Easy-to-miss costs beyond text tokens
Tools such as web search, file search and hosted containers can add charges beyond the basic text-token estimate. Image, audio and realtime workloads have their own pricing dimensions. Regional processing can also change the applicable rate. Check the official pricing documentation and the provider invoice before setting a production budget. Start with a measured sample of real requests, then include retries and multi-step runs in your forecast.
How to use this data
- Recalculate with your real input/output token mix.
- Check region, rate limits, and payment method.
- Recheck official sources before purchasing or switching.
Frequently asked questions
How much does 1 million GPT tokens cost?
There is no single GPT token price. Select the exact model and channel, then distinguish input from output and any cache or context tier. The table shows tracked input and output rates per million tokens in each channel’s billing currency.
Can I use a ChatGPT subscription for API-key calls?
API-key usage follows API billing, not the included ChatGPT subscription allowance. Subscription access to Codex is a separate usage path; check which authentication method your tool uses before estimating cost.
Is the lowest input rate enough to compare total cost?
No. Multiply both input and output rates by your measured token volumes, then add applicable tool and service fees. A low input rate alone cannot establish the lowest total cost, and a missing output rate is unknown, not zero.
Does this comparison include every OpenAI model and fee?
No. It displays up to 12 recently listed, tracked GPT-family language models with available per-million-token channel prices. It is not a complete catalog and does not estimate image, audio, storage or tool charges. Use the official rate card for those products.
Official sources and review date
Billing guidance reviewed: . Price verification dates are listed separately in the table.