OpenAI slashes GPT-5.6 Sol costs as AI pricing competition intensifies
The frontier model in OpenAI’s latest family sees input costs drop by 20 per cent and output costs by 33 per cent under a new promotional structure.
OpenAI has moved to reduce the cost of access to its GPT-5.6 Sol model, the flagship offering within its latest GPT-5.6 family. The update, detailed in the company’s API documentation, sets the new rate at $4 per million input tokens and $20 per million output tokens. This adjustment represents a 20 per cent reduction in input pricing and a 33 per cent reduction in output pricing compared to previous rates.
The company has designated these new rates as promotional pricing, with the structure available at least through 21 November 2026. While the specific duration of the promotion may extend beyond this date, the explicit labelling suggests these figures are not yet the permanent standard rate for the model.
GPT-5.6 Sol corresponds to the unsuffixed model tier used in earlier GPT-5 families, serving as the primary entry point for developers seeking the most advanced capabilities in the current generation. The `gpt-5.6` alias is configured to route requests directly to this specific model, ensuring that users accessing the base identifier are directed to the Sol variant.
For developers working with large context windows, the pricing structure includes specific multipliers. Prompts exceeding 272,000 input tokens are billed at 2x the standard input rate and 1.5x the standard output rate for the full request. Additionally, cache writes are charged at 1.25x the uncached input token rate, a detail that may impact cost predictability for high-volume institutional users.
The model also supports a range of reasoning effort settings, including none, low, medium, high, xhigh, and max, with medium set as the default. This flexibility allows users to balance computational cost against the depth of inference required for their specific applications.
The price cut lands as OpenAI continues to refine its positioning in a competitive artificial intelligence landscape. By lowering the barrier to entry for its frontier model, the company aims to attract more developers and enterprises to its ecosystem while maintaining a clear distinction between its various model tiers.
