OpenAI slashed the price of its flagship GPT-5.6 Sol model by more than 20 percent on August 21, marking the third price cut for the GPT-5.6 family in under a month and the first reduction for its most capable tier. The move places Sol below rival offerings from Anthropic on both input and output pricing, signaling intensifying competition in the frontier model market.
The New Prices
OpenAI dropped GPT-5.6 Sol's input cost from $5 to $4 per million tokens and output from $30 to $20 per million tokens for standard short-context use. Cached input fell from $0.50 to $0.40 per million tokens, and cached output from $0.50 to $0.40 as well. The promotional rate is available through at least November 21, 2026, and applies to the pay-as-you-go API, Codex credits, and eligible ChatGPT Work plans. Consumer subscription tiers -- Pro, Plus, and Business -- remain at their current rates.
The announcement came through the OpenAI Developer Community, posted on August 21, where the company wrote that the reduction reflects "pushing the frontier of capabilities while improving efficiency."
A Pattern of Rapid Repricing
This is the second wave of GPT-5.6 pricing cuts. On July 30, OpenAI cut Luna -- its lowest-tier model -- by 80 percent, bringing it to $0.20/$1.20 per million tokens for input and output. Terra followed with a 20 percent cut to $2/$12 per million. Sol held its line through those changes, positioned as the premium tier with no sign of softening. Three weeks later, that changed.
AWS confirmed the pricing change in an announcement posted on August 21, noting that Sol is now available through Bedrock at the new rates. The speed of this repricing -- three tiers cut within weeks -- is unusual for a model line that was just launched.
The Competitive Context
OpenAI's price cut directly undercuts Anthropic's Claude Opus 5, which lists at $5 per million input tokens and $25 per million output tokens. Sol is now cheaper than Opus 5 on both input and output, a notable shift given that Sol has been the more expensive option since launch. Anthropic's Claude Fable 5, priced at $10/$50 per million tokens, is also more expensive across the board.
The pressure is coming from multiple directions. Google's Gemini 3.7 Flash launched recently with stronger agent performance at roughly half the price of the previous generation. Anthropic has been chipping away at enterprise accounts by emphasizing safety tooling and reliability. Chinese AI labs including DeepSeek and Moonshot AI have been producing increasingly capable models at aggressive prices.
For enterprises running agentic workflows where a single task involves hundreds or thousands of model calls, the difference between $5 and $4 per million input tokens, or $30 and $20 output, compounds quickly across production workloads.
What This Means for Developers
The price cut requires no code changes or migration. Existing integrations with the GPT-5.6 Sol API will automatically reflect the new rates. For teams running high-volume inference -- autonomous document review, multi-step agent workflows, complex coding tasks -- the lower output pricing makes previously marginal use cases economically viable.
Enterprise customers on negotiated agreements should check whether their contracts tie pricing to the public rate, as some agreements may not automatically track promotional periods.
The Bigger Pattern
Three data points make a clear trend. OpenAI cut Luna by 80 percent, Terra by 20 percent, and now Sol by more than 20 percent, all within a few weeks. This is not routine optimization; it is a deliberate repricing of the entire GPT-5.6 family to respond to competitive pressure.
The frontier model business, which looked like a high-margin premium market six months ago, is behaving more like a commodity infrastructure race. For businesses building on top of these models, the window to do ambitious work cheaply is opening wider every week. For AI lab valuations, the race to the bottom on pricing may be less welcoming.
The next question is whether other labs will match these cuts or hold firm. Anthropic's recent launch of Claude Opus 5 positioned the company as a safety-first alternative, but price is becoming a factor in every category. If OpenAI continues this trajectory, the frontier model market could see another round of significant price movement before November, when the promotional period ends.