Short answer: An excerpt from OpenAI’s GPT-4 model page shows $30 per 1 million input tokens and $60 per 1 million output tokens next to the label “Batch API price.” The available excerpt does not establish whether that label applies to both amounts, whether they are standard or Batch API rates, or which currency the dollar sign denotes. Treat them only as figures in an ambiguously labeled pricing display—not as confirmed general GPT-4 API rates. The same excerpt marks GPT-4 and the listed gpt-4, gpt-4-0613, and gpt-4-0314 snapshots as shut down, so these figures are not a production quote for sending new GPT-4 requests. (OpenAI GPT-4 model page)
The pricing information available for this article was reviewed on November 9, 2026. The excerpt itself is undated, and the page could not be re-retrieved to confirm that the display remains unchanged. That review date is not a price effective date.
Hypothetical calculation using the displayed amounts
If you use the two amounts purely as arithmetic inputs, the calculation is:
(input tokens ÷ 1,000,000 × $30) + (output tokens ÷ 1,000,000 × $60)
For example, a hypothetical workload of 100,000 input tokens and 20,000 output tokens gives:
- Input: 100,000 ÷ 1,000,000 × $30 = $3.00
- Output: 20,000 ÷ 1,000,000 × $60 = $1.20
- Total: $4.20, using the amounts shown in the excerpt
This is only an illustration of the arithmetic. Because the displayed channel label and currency are unresolved—and GPT-4 is marked shut down—the result is not a verified estimate of a standard API bill, a Batch API bill, or a new production workload.
Token counts are not word counts. The tokens sent to a model contribute to input usage, while generated tokens contribute to output usage; the full request structure and attached content can affect token counts. (OpenAI Help Center: Understanding and counting tokens)
What to use for a production budget
For a new production integration, do not budget on the assumption that GPT-4 is available. Check the current pricing and availability for the exact active model and API path you intend to use. The figures above do not establish a current GPT-4 production rate.
Keep API charges separate from ChatGPT subscriptions: a subscription price is not an API token rate. (OpenAI Help Center: Managing billing for ChatGPT and the API platform)
Bottom line: The available GPT-4 page excerpt displays $30/M input and $60/M output beside “Batch API price,” but its channel scope and currency are unresolved. GPT-4 is marked shut down. Treat the numbers as an ambiguous display, not a confirmed price for a production API budget.