OpenAI is sharply lowering the cost of two GPT-5.6 models, with the biggest change aimed at its most affordable option. The company is cutting GPT-5.6 Luna prices by 80 percent and GPT-5.6 Terra prices by 20 percent, effective July 30.
The move changes the economics for developers and teams using OpenAI models through ChatGPT Work, Codex, and the OpenAI API. It also lands in a market where price pressure is rising, especially from low-cost Chinese providers.
What changed in GPT-5.6 pricing
The largest reduction is for GPT-5.6 Luna. Luna drops to $0.20 per million input tokens and $1.20 per million output tokens. That puts the model in a much lower cost bracket for workloads that involve high token volume.
GPT-5.6 Terra is also getting cheaper, though by a smaller margin. Terra falls to $2 per million input tokens and $12 per million output tokens after a 20 percent cut.
OpenAI is not changing GPT-5.6 Sol pricing. In this update, Luna and Terra carry the price reductions, while Sol remains at its current level.
- GPT-5.6 Luna: 80 percent price cut, now $0.20 per million input tokens and $1.20 per million output tokens.
- GPT-5.6 Terra: 20 percent price cut, now $2 per million input tokens and $12 per million output tokens.
- GPT-5.6 Sol: pricing stays the same.
Why Luna matters for cheaper AI workloads
OpenAI says Luna matches the performance of leading models from a year ago. That comparison is important because many AI applications do not always need the most expensive frontier option. They need a model that is capable enough, fast enough, and priced low enough to use repeatedly.
According to OpenAI, a task that cost a dollar with those earlier leading models now runs about 6 cents on Luna. The company also says Luna is nearly nine times faster. For teams building products around large volumes of requests, those two claims point to a very different cost profile.
The practical implication is simple: lower prices can expand the set of jobs where an AI model is financially reasonable. Tasks that previously looked too expensive at scale may become easier to justify when the input and output costs fall this far.
This does not mean Luna replaces every higher-end model. The source only says Luna matches the performance of leading models from a year ago, while Sol pricing stays the same. But it does suggest OpenAI wants a stronger low-cost option inside the GPT-5.6 lineup.
Infrastructure gains are part of the explanation
OpenAI says the cuts are possible because GPT-5.6 Sol made the company's own infrastructure more efficient. The model allegedly optimized GPU software on its own, cutting deployment costs by 20 percent.
The company also points to better token generation. OpenAI says token generation improved by more than 15 percent through speculative decoding.
Those details matter because model pricing is tied to the cost of running models at scale. If infrastructure becomes cheaper or faster, providers have more room to lower prices, protect margins, or both. In this case, OpenAI is presenting the price cuts as connected to internal efficiency gains rather than only external competition.
At the same time, the source notes that market pressure likely played a role too. Lower operating costs may create the ability to cut prices, while competitive pressure can create the reason to do it now.
The competitive signal behind the cuts
The AI market is facing growing price pressure, especially from low-cost Chinese providers. That pressure changes how model companies compete. Performance still matters, but price becomes harder to separate from the overall product strategy.
Microsoft is also now openly promoting its own MAI models as cheaper alternatives to OpenAI. That adds another layer to the pricing environment, because OpenAI is not only competing with outside providers. It is also being compared against cheaper model options promoted by a major technology partner.
The source describes this as a price war that could hurt the broader market if it slows revenue growth at frontier labs. The concern is that these labs have balance sheets tied to massive infrastructure investments. If prices fall quickly while infrastructure costs remain large, the financial pressure can increase.
For customers, cheaper access can be immediately useful. For model providers, the same trend creates a harder business problem: they must keep improving capability, serve more usage, and manage infrastructure spending while customers expect lower prices.
What users can take from the update
For users of ChatGPT Work, Codex, and the OpenAI API, the important point is availability. All GPT-5.6 models are available through those channels, and the new Luna and Terra pricing changes the cost calculation for using them.
Luna is the clearest story in this update. Its 80 percent cut, $0.20 per million input tokens, and $1.20 per million output tokens make it the model most directly aimed at affordability. Terra's 20 percent reduction gives users a cheaper option at a higher price point than Luna, while Sol stays unchanged.
The larger takeaway is that AI model pricing is becoming more aggressive. OpenAI is lowering costs, citing infrastructure improvements and operating gains, while also facing a market where lower-cost alternatives are increasingly visible. That combination makes price a central part of the GPT-5.6 story, not a detail on the side.