OpenAI has announced a large-scale reduction in the cost of using the GPT-5.6 model family, including a fivefold price cut for the Luna model and the launch of a high-speed "Fast" mode for the flagship Sol model.
What Happened
The cost of the Luna model has decreased 5x and now stands at $0.2 per 1 million input tokens and $1.2 per 1 million output tokens. The price of the Terra model has dropped by 20%, reaching $2 per 1 million input tokens and $12 per 1 million output tokens. Additionally, a "Fast" mode has been introduced for the flagship Sol model, providing speeds 2.5x faster than standard at double the cost.
Context
This update to pricing and functionality comes amid active development of infrastructure optimizations aimed at reducing latency and increasing inference efficiency for large-scale systems.
Why It Matters for the Industry
The sharp reduction in the cost of the fastest models (Luna) stimulates the mass adoption of AI agents and high-frequency LLM tasks where the price per million tokens is critical. This intensifies price competition in the efficient model segment and puts pressure on manufacturers of Small Language Models (SLMs).
Why It Matters for Users
Development is becoming significantly more affordable. Using the inexpensive Luna models allows for the construction of more complex, large-scale, and multi-agent systems, offloading routine tasks from heavy flagship models to economically efficient solutions without significant loss of quality in typical scenarios.
Sources
Author
Look at AI, Editorial Team