OpenAI quietly lowered the price of its flagship GPT-5.6 Sol model by more than 20% on August 21, 2026, marking the second price reduction in under a month for the GPT-5.6 family and continuing what is now an unmistakable trend: frontier AI capabilities are getting cheaper, faster than most enterprises planned for.

The new pricing drops Sol’s input cost from $5 to $4 per million tokens, and output from $30 to $20 per million tokens. The promotional rate runs through at least November 21, 2026, and applies to the pay-as-you-go API, Codex credits, and eligible ChatGPT Work plans. Consumer subscriptions like Pro, Plus, and Business stay at their existing prices.

The timing matters. When our previous coverage noted OpenAI’s Luna and Terra cuts in late July, Sol was explicitly left unchanged. Sol was positioned as the expensive, full-capability tier with no signs of softening. That changed in three weeks.

The Competitive Context

This cut directly undercuts Anthropic’s Claude Opus 5 on both input and output pricing. For enterprises building agentic workflows where a single task can involve hundreds or thousands of model calls, the difference between $5 and $4 input, or $30 and $20 output, compounds fast across a production workload.

The background pressure here is straightforward: Google’s Gemini 3.7 Flash launched this week with stronger agent performance at roughly half the price of the previous generation, and Anthropic has been chipping away at enterprise accounts by emphasising reliability and safety tooling. OpenAI is responding with price, which is not unusual at this stage of a competitive market cycle.

What is unusual is the speed. Luna took 80% off its price tag just three weeks after launch. Sol held for longer, but not by much. The frontier model business, which looked like a high-margin premium game six months ago, is behaving more like a commodity infrastructure race.

What This Means for Business

If you’re already using GPT-5.6 Sol in production, your API bill just dropped. No code changes, no migration work, just a lower invoice. If you’re on an enterprise agreement rather than pay-as-you-go, check whether your contract ties pricing to the public rate.

If you’ve been holding off on frontier model use cases because of cost, the calculus has shifted. Tasks that required Sol-level reasoning but felt economically marginal at $30 output per million tokens may now make sense at $20. Autonomous document review, complex multi-step agent workflows, high-accuracy coding tasks for internal tools, these all get more viable as output pricing drops.

If you’re a decision-maker evaluating AI vendors, this is a good moment to go back to your model usage data. If your organization is sitting on a small model because a large one “costs too much,” the assumptions from your last review may no longer hold.

The Bigger Pattern

Three data points make a trend. OpenAI cut Luna by 80%, Terra by 20%, and now Sol by 20-plus percent, all within a few weeks of each other. That is not routine optimisation, that is a deliberate repricing of the entire model family to respond to competitive pressure.

For enterprises, this is broadly good news. More capability per dollar means you can do more with the same budget, or spend the same and get significantly more throughput. The risk to watch is the speed of change itself: pricing assumptions built into AI ROI models earlier this year may already be stale.

At Enterprise DNA, we work with organisations across industries on how to structure AI adoption so it scales efficiently, not just what models to use. If your team is trying to figure out which models actually make sense for your workloads at current pricing, that is exactly the kind of question our Omni Advisory service exists to answer. The model market is moving fast, and the right answer three months ago is not necessarily the right answer today.

The race to the bottom on frontier model pricing may not be great for AI lab valuations, but for businesses building real workflows on top of these models, the window to do ambitious things cheaply is opening wider every week.