OpenAI Cuts GPT-5.6 Luna Price by 80 Percent to 20 Cents Per Million Tokens

OpenAI cut the price of its GPT-5.6 Luna model by 80 percent in August 2026, dropping the cost to $0.20 per million input tokens. The reduction makes Luna one of the most affordable frontier AI models available to developers.

The price cut reflects intensifying competition among AI providers. Alibaba's Qwen3.8-Max, released in early August, features 2.4 trillion parameters and a 1-million-token context window. DeepSeek launched V4-Flash, positioning it as the lowest-cost model on key benchmarks.

Anthropic's Claude Opus 5 supports a 1-million-token context window, and Claude Fable 5 scored 80.3 percent on the SWE-Bench Pro coding benchmark. Google released Gemini 3.0 and 3.6 Flash, emphasizing real-time world modeling and spatial audio and video processing.

The rapid price declines are reshaping how businesses use AI. Tasks that were cost-prohibitive a year ago, such as processing millions of customer service interactions or analyzing large document sets, are now economically viable for mid-sized companies.

OpenAI also introduced GPT-5.6 in three tiers: Luna, Terra, and Sol. The company said the tiers are designed to give developers options based on their cost and performance needs.

The company's internal System 2 architecture for GPT-5.5 allows models to allocate compute for logical planning before generating responses, reducing hallucinations in complex tasks.

AI infrastructure costs remain high despite falling model prices. A proposed $100 billion AI computing campus in Kentucky, featuring 2 gigawatts of gas generation and 2.6 gigawatts of battery storage, illustrates the scale of investment required to run frontier models.

Developers and businesses are being advised to test multiple models before committing to a provider, as the competitive landscape is shifting rapidly.