GPT-5.5 fine-tune costs $2,325
- OpenAI’s current API pricing implies a GPT-5.5 run with 11.52 million long-context input tokens would cost about $115.20 before any output, caching, or surcharges. - The $2,325 figure matches long-context GPT-5.5 pricing only if the job also generated roughly 49.1 million output tokens at $45 per million. - OpenAI now lists GPT-5.5 at $5 input, $0.50 cached input, and $30 output per million for short context, with higher long-context rates. (openai.com)
OpenAI’s published API pricing does not support a flat claim that “11.52 million GPT-5.5 fine-tune tokens cost $2,325” on input alone. At current rates, that input volume is far cheaper unless the run also produced a very large number of output tokens. (openai.com) (developers.openai.com) OpenAI’s pricing page lists GPT-5.5 at $5 per million input tokens, $0.50 per million cached input tokens, and $30 per million output tokens for shorter prompts. Its detailed pricing table lists higher long-context rates of $10 input, $1 cached input, and $45 output per million. (openai.com) (developers.openai.com) Using the long-context rates in that table, 11.52 million input tokens would cost $115.20 if none were cached. If every one of those tokens were billed as cached input instead, the charge would be $11.52. (developers.openai.com) To reach a total near $2,325 at those same long-context rates, the run would need about $2,209.80 in output charges. At $45 per million output tokens, that works out to roughly 49.1 million output tokens. (developers.openai.com) That distinction matters because OpenAI bills input, cached input, and output separately. A cost screenshot that mixes those categories can make a training or tuning run look far more expensive than the raw input-token count suggests. (developers.openai.com 1) (developers.openai.com 2) OpenAI’s docs also say GPT-5.5 supports prompt caching, while GPT-5.5 Pro does not offer a cached-input discount. The Pro model is priced much higher, at $30 per million input tokens and $180 per million output tokens in batch mode, with no cached-input line item. (developers.openai.com 1) (developers.openai.com 2) OpenAI’s model docs describe GPT-5.5 as its flagship model for complex reasoning and coding, and the company said on April 23, 2026 that GPT-5.5 and GPT-5.5 Pro were rolling out to ChatGPT first, with API access coming “very soon.” That timing makes screenshots and third-party breakdowns especially easy to misread while pricing pages and product rollouts are still settling. (developers.openai.com) (openai.com) The clean read is narrower than the viral number: 11.52 million GPT-5.5 input tokens do not equal a $2,325 bill by themselves. That total only pencils out if the job included tens of millions of billed output tokens, a pricier tier, or both. (developers.openai.com)