OpenAI is making its most capable GPT-5.6 model cheaper at a moment when developers are starting to judge frontier AI on more than intelligence alone. GPT-5.6 Sol will temporarily cost $4 per million input tokens and $20 per million output tokens, reducing the price of running complex reasoning, coding and agentic workloads through OpenAI’s flagship model.
The change represents a 20% reduction in input pricing and roughly a 33% reduction in output pricing from the previous $5 and $30 rates. The lower prices are expected to remain in place for three months and apply to API usage, with related pricing changes also rolling out across eligible credit-based usage in ChatGPT Work and Codex. Consumer subscription prices for Plus, Pro and Business remain unchanged.
The timing makes the move more interesting than the discount itself.
OpenAI introduced GPT-5.6 as a family built across three economic tiers. Sol handles the most demanding professional work, Terra balances capability with cost, and Luna targets high-volume workloads. Terra and Luna had already received substantial price reductions in July, while Sol’s standard pricing remained unchanged at that point.
Now the frontier tier is being pulled into the same efficiency race.
That matters because advanced models increasingly sit inside software that makes thousands or millions of calls rather than occasional chatbot conversations. Coding agents, research systems and automated enterprise workflows can generate long outputs, making token economics a direct product-design constraint.
A reduction from $30 to $20 per million output tokens therefore changes more than an API bill. It increases the number of tasks where developers can justify using the strongest model instead of routing work toward a cheaper alternative.
The next phase of model competition may be decided as much by cost per successful task as benchmark performance. OpenAI’s temporary Sol pricing gives developers three months to test exactly how much more frontier intelligence they can afford to put into production.
