OpenAI slashes Luna prices and accelerates Sol in the API

OpenAI slashes the price of its Luna model by 80% and adds a fast speed option for Sol. The impacts of this update on your API budgets.

The GPT-5.6 lineup is revising its pricing structure. OpenAI is cutting the price of Luna, its fastest and least expensive model, by 80%, while Terra benefits from a 20% reduction. The new API rates are set at $0.20 per million input tokens and $1.20 per million output tokens for Luna, compared to $2.00 input and $12.00 output for Terra.

These price drops also modify quota consumption in Codex and ChatGPT Work. Subscriptions and their limits remain unchanged, but using Luna and Terra now consumes fewer credits. OpenAI is also upgrading the Auto-review feature from GPT-5.4 to GPT-5.6 Luna in the ChatGPT application and the Codex CLI. The company estimates that this combination can reduce its cost by approximately tenfold.

For more demanding tasks, GPT-5.6 Sol retains its standard pricing but gains a Fast mode in the API. This option promises up to 2.5 times faster generation, with no announced changes to the model's capabilities, for double the price. It replaces Priority Processing, while requests already configured with the `priority` parameter will automatically switch to this new processing method.

OpenAI attributes part of these savings to optimizations made across its entire infrastructure. GPT-5.6 Sol reportedly assisted in rewriting components executed on GPUs, improving its own generation system, and fine-tuning production configurations. According to the company, this work has reduced the overall running cost by 20% and improved generation efficiency by more than 15%.

The distinction between the three models thus becomes clearer: Luna targets high volumes and well-defined tasks, Terra is for everyday use cases requiring more versatility, and Sol is for complex problems where the level of reasoning or speed justifies a higher cost.