GPT API cost depends on the selected GPT model and the number of normal input, cached input, and output tokens. Multiply each token group by its published per-million rate, then add the results. Use the calculator below for a workload estimate and confirm the current OpenAI source before funding or deploying an application.
This estimate uses published text-token rates. It excludes tool calls, media, storage, taxes, regional pricing, platform fees, and provider-side token differences.
The GPT API cost calculator answer depends on the same input, cached input, and output token counts. Compare the listed OpenAI rates, calculate each shortlisted model with one representative workload, and do not treat a lower token price as proof of better task value. Missing prices remain unavailable rather than becoming zero.
| Provider / model | Input / 1M | Cached / 1M | Output / 1M | Context | Accuracy | Verified |
|---|---|---|---|---|---|---|
| GPT-5.4 miniOpenAI | $0.75 | $0.075 | $4.5 | 400,000 | Compatible | 2026-08-09Verified |
| GPT-5.4 nanoOpenAI | $0.2 | $0.02 | $1.25 | 400,000 | Compatible | 2026-08-09Verified |
| GPT-5.6 LunaOpenAI | $0.2Over 272K: $0.4 | $0.02Over 272K: $0.04 | $1.2Over 272K: $1.8 | 1,050,000 | Compatible | 2026-08-09Verified |
| GPT-5.6 SolOpenAI | $5Over 272K: $10 | $0.5Over 272K: $1 | $30Over 272K: $45 | 1,050,000 | Compatible | 2026-08-09Verified |
| GPT-5.6 TerraOpenAI | $2Over 272K: $4 | $0.2Over 272K: $0.4 | $12Over 272K: $18 | 1,050,000 | Compatible | 2026-08-09Verified |
The 17 model rows use USD list prices per one million text tokens, checked against the linked official provider pages. OpenAI and Gemini long-context tiers are applied automatically from the entered input count. Batch, audio, image, video, tools, storage, regional, and platform fees may differ.
Normal input, eligible cached input, and generated output can use different rates. The GPT API cost calculator caps cached input at total input and prices unsupported cache scenarios at the normal input rate.
Message framing, tools, media, and endpoint behavior can change provider usage beyond the text you counted locally.
The GPT API cost calculator multiplies cost per request by requests per day and active days per month. Run a typical case, a long-output case, and a retry case instead of trusting one optimistic workload.
The GPT API cost calculator applies the published cached-input rate only to the cached token amount entered. The GPT API cost calculator does not infer cache writes, minimum size, retention, or hit rules; those still follow OpenAI's current documentation.
The GPT API cost calculator covers OpenAI models. Open the GPT vs Claude API cost comparison for a side-by-side estimate, then review all models in LLM API pricing.
The GPT API cost calculator measured normal input, cached input, output, zero-token, 1-token, 1 million-token, 10 million-token, and high-volume cases against the centralized OpenAI rates. Its limitation is billing scope: tools, media, retries, storage, regional modes, taxes, and provider-side tokenization remain outside the estimate. OpenAI official API pricing. Last verified: .
| Evidence | Observed result or boundary |
|---|---|
| Measured GPT cases | 0, 1, 1,000, 1 million, and 10 million tokens with cache and output variations. |
| Comparison boundary | GPT results can be compared with Anthropic Claude, Google Gemini, and DeepSeek rates. |
GPT API cost questions
The GPT API cost calculator adds normal input cost, cached input cost, and output cost. Each group uses the selected model's published USD rate per one million tokens.
No. The GPT API cost calculator models text-token list prices. Tools, images, audio, storage, regional processing, taxes, and platform fees can add separate cost.
No. The GPT API cost calculator applies the cached-input rate only to the amount entered; it does not prove that a request qualifies for or hits a provider cache.
No. Use the GPT API cost calculator beside task quality, output length, retries, latency, context, tools, privacy, and operational fit.
The GPT API cost calculator is strongest when input, output, cache, and volume come from representative requests. It is not for invoicing; provider usage remains the billing source of truth.