--- title: "OpenAI Strikes Hard! GPT-5.6 Series Prices Plunge Up to 80%, AI Begins Self-Optimizing Code for Instant Leap" type: "News" locale: "en" url: "https://longbridge.com/en/news/294433925.md" description: "OpenAI announced significant price cuts for the GPT-5.6 series, with the Luna model's price dropping by 80% and Terra by 20%. Code inspection costs have decreased tenfold. The Sol model introduces a Fast Mode, boosting speed by 2.5 times. The new prices take effect on July 30, aiming to accelerate the arrival of an era of affordable, unlimited AI usage" datetime: "2026-07-31T00:58:48.000Z" locales: - [zh-CN](https://longbridge.com/zh-CN/news/294433925.md) - [en](https://longbridge.com/en/news/294433925.md) - [zh-HK](https://longbridge.com/zh-HK/news/294433925.md) --- # OpenAI Strikes Hard! GPT-5.6 Series Prices Plunge Up to 80%, AI Begins Self-Optimizing Code for Instant Leap OpenAI suddenly announced comprehensive price cuts for the GPT-5.6 series today. The price of Luna, the fastest and most affordable model, has plummeted by 80%. The price of Terra, designed for daily office tasks, has been reduced by 20%. Additionally, the cost of the auto-approval mode in applications, which helps check code, has dropped tenfold. Furthermore, Sol, the model focused on highest intelligence, has launched a Fast Mode. Its speed has surged by up to 2.5 times. According to OpenAI, the era of truly unlimited and extremely cheap AI usage is accelerating. ## Price Cut Details Let's look at the price table first; I've compiled some statistics. Sol's price remains unchanged, but its speed has been individually enhanced, which refers to the Fast Mode discussed below. The input price for DeepSeek V4 Pro is the quote when there is a cache miss; if there is a cache hit, the price drops to $0.003625 per million tokens. Looking at intelligence level and cost together provides a clearer picture. Third-party evaluation firm Artificial Analysis gave DeepSeek V4 Pro (Max reasoning mode) an intelligence index score of 44.27, with a cost per task of approximately $0.0448. Converting this to a cost-to-intelligence ratio, DeepSeek V4 Pro is 1.9 times higher than GPT-5.6 Luna High mode, meaning Luna High is more cost-effective for the same level of intelligence. The new prices take effect on July 30 and will subsequently be synchronized to the AWS platform. Subscription prices and quotas for ChatGPT and Codex remain unchanged, but using Terra and Luna for tasks will consume fewer quota units. ## A New Feature: Fast Mode OpenAI has added a Fast Mode to its API to replace the previous priority processing feature. For Sol, the speed in Fast Mode can reach up to 2.5 times that of the standard mode, at twice the price of the standard mode, with no change in intelligence level. Fast Mode is backward compatible; requests previously marked for priority processing will automatically switch to Fast Mode without requiring additional user action. In Codex, this feature corresponds to the /fast mode. ## Auto-Approval Mode, Costs Drop Tenfold The cost of the auto-approval feature for agents (known as "Review for Me" in the app) has dropped to about one-tenth of its original cost. This feature now uses the Luna model to intercept high-risk operations in advance, preventing the main agent from directly executing potentially problematic actions. ## The Core Secret Behind the Price Cuts: AI Begins to Optimize Itself How were such significant price cuts and speed improvements achieved? The core secret lies in the models beginning to optimize themselves. **Under the guidance of human engineers, the Sol model autonomously rewrote and optimized the underlying production kernel. It designed and ran hundreds of experiments to improve token generation efficiency, and even monitored model training, intervening directly to resolve issues when they arose.** This approach directly reduced the model's end-to-end operating costs by 20% and increased token generation efficiency by 15%. The underlying routing has become smarter, software generates tokens more efficiently, and context management allows AI to avoid redundant work. The same computing power can now accomplish more. This creates a positive feedback loop: the stronger the model, the more it can autonomously participate in optimization efforts, which in turn makes the next generation of models cheaper and more efficient. ## Who Can Access These Price Cuts Terra and Luna are available in ChatGPT Work, Codex, and the OpenAI API. Free and Go users can use Terra, while Plus, Pro, Business, and Enterprise users can choose freely between Terra and Luna. ## Why Do This OpenAI places these price cuts within a broader strategy: matching the appropriate model based on the task's importance, cost of error, urgency, and scale. For example, in a typical programming workflow, one can first use Sol to clarify requirements and define the solution, then hand off to Luna to write code, run tests, and verify results. This approach lowers overall costs while ensuring quality. OpenAI also stated that it will continue to invest in both stronger computing power and more efficient ways to utilize it. One path involves making large-scale tasks cheaper with models like Luna and Terra, while the other involves using Sol with Fast Mode to run the most critical tasks faster. Risk Warning and Disclaimer The market carries risks; investment requires caution. This article does not constitute personal investment advice, nor does it consider the specific investment objectives, financial status, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article align with their specific circumstances. Investment decisions made based on this content are the sole responsibility of the investor. ### Related Stocks - [OpenAI.NA](https://longbridge.com/en/quote/OpenAI.NA.md) ## Related News & Research - [OpenAI blinks in face-off with Chinese rivals, making big price cuts](https://longbridge.com/en/news/294528211.md) - [Meet the Gen Z AI whiz at the center of a hedge fund meltdown](https://longbridge.com/en/news/294425537.md) - [OpenAI Cuts Price Of GPT-5.6 Luna By 80% & GPT-5.6 Terra By 20%](https://longbridge.com/en/news/294396695.md) - [In the Hugging Face breach, OpenAI’s hacker was noisy and fast — but not unstoppable](https://longbridge.com/en/news/294384788.md) - [OpenAI CFO Sarah Friar tells employees that annualized revenue in July topped all of Q2](https://longbridge.com/en/news/294273022.md)