OpenAI GPT-5.6 Luna Price Slashed by 80%, AI Inference Costs Plummet
One-sentence hook
OpenAI slashes GPT-5.6 Luna prices by 80%, bringing input tokens to just $0.20 per million and ending the "AI as luxury good" era.
## Three Key Information
-
Significant Price Reduction: OpenAI announces a dramatic reduction in GPT-5.6 Luna model input token pricing from $1.00 per million to just $0.20—a whopping 80% cut. Output token prices are also adjusted downward, making high-frequency AI calls extremely affordable for SMEs and independent developers.
-
Daybreak Program Upgrade Launches Simultaneously: Alongside the price adjustment, OpenAI upgrades its Daybreak cybersecurity initiative, introducing GPT-5.6-Cyber, a model designed specifically for defenders that completed 95% of advanced cybersecurity tasks in testing. This marks OpenAI's expansion of AI from general domains to critical infrastructure protection.
-
Market Response and Intensified Competition: Alibaba's Qwen3.8-Max follows as an open-source competitor, narrowing the performance gap between open and closed-source models. Analysts say this price war will further accelerate AI adoption in enterprise applications, especially in scenarios requiring massive token processing, where cost advantages will become decisive factors.
## WangDou's Take
This move is nothing short of a "diving strike." OpenAI's 80% price cut directly slashes AI inference costs from "rich people's toy" back into the "mass market." Remember, GPT-4's input token price was several times higher previously—now Luna at $0.20 essentially opens AI large models to SMEs, indie developers, and even individuals. This isn't pure benevolence—it's survival. The AI industry is burning money like water; those who don't cut prices get eliminated. Moreover, this price war directly fuels open-source alternatives like Qwen, and everyone is swimming naked in the price war. Ultimately, the real beneficiaries are ordinary users who need AI. Just be cautious: cheaper doesn't always mean better. With lower token prices, model capability and privacy compliance may warrant closer inspection.
Source: Bloomberg