GPT-5.6 Sol API prices
OpenAI’s GPT-5.6 Sol model documentation lists these token prices:
| Token type | Listed price |
|---|---|
| Input | $4.00 per 1 million tokens |
| Cached input | $0.40 per 1 million tokens |
| Output | $20.00 per 1 million tokens |
The page says promotional pricing is available through November 30, 2026. It does not state what prices will apply after that date. These are rates listed on the OpenAI model page—not a subscription price or a confirmed TokenForge Cloud, reseller, or gateway price. The page displays dollar signs but does not establish currency or channel-specific terms, so confirm the applicable rate with your API provider before committing a budget.
Estimate your token cost
For a request without long-prompt pricing or other charges, a basic estimate is:
(uncached input tokens ÷ 1,000,000 × $4) + (cached input tokens ÷ 1,000,000 × $0.40) + (output tokens ÷ 1,000,000 × $20)
For example, 200,000 uncached input tokens and 100,000 output tokens come to $2.80 at the listed rates: $0.80 for input plus $2.00 for output. This illustration assumes the request does not exceed the long-prompt threshold and has no cache-write or tool-call charges.
Pricing rules that can change the estimate
- Long prompts: For a prompt with more than 272,000 input tokens, the page says input is priced at 2× and output at 1.5× for the full request. For example, 300,000 input tokens and 100,000 output tokens would calculate to $2.40 for input and $3.00 for output, or $5.40 total, using the listed rates and assuming no other charges.
- Cache writes: The page says cache writes are billed at 1.25× the uncached input rate. At the listed $4 rate, that is $5 per million cache-write tokens. This is distinct from the $0.40-per-million cached-input rate; don’t treat the two as the same charge.
- Tool calls: The page notes that some tool-specific models, including search and computer use, charge per tool call. It does not give the exact fees, so include them separately in a production estimate when applicable.
These calculations use the rates shown on the model page and are not all-in production quotes. The page does not explain how the long-prompt or cache-write rules interact with the promotional end date.
Production API fit: what to check
The model page identifies the model as gpt-5.6-sol and says the gpt-5.6 alias routes to it. It lists a 1,050,000-token context window, a 128,000-token maximum output, text and image input, and text output. Streaming, function calling, and structured outputs are listed as supported.
The page marks audio and video as unsupported, even though it also lists audio- and video-named endpoints. That discrepancy is not explained in the captured information: don’t assume those endpoints mean GPT-5.6 Sol supports audio or video. Confirm endpoint compatibility and any tool charges in the relevant official documentation before implementation.
For a budget decision, use the listed token rates as a starting point, then account for prompt size, cached input and cache writes, any billable tool calls, and the API provider or channel you will actually use. No post-promotion price is stated on the model page.