On September 22, 2026, OpenAI added Sol and Luna to its GPT-6 product line. These models were designed to make some of the technical improvements from the previously released high-performance model, Astra, available at a lower cost.
API pricing for GPT-6 Sol is $2 per 1 million input tokens and $10 per 1 million output tokens, while GPT-6 Luna is $0.10 and $0.50, respectively. More important than the mere announcement of a new model is the fact that users can now choose a model that fits their workload and budget.
| How much will the price differ from the previous generation?
According to OpenAI's official announcement, the promotional price for GPT-5.6 Sol was $4 per input and $20 per output. GPT-6 Sol has been reduced by half in each case. GPT-5.6 Luna cost $0.20 for input and $1.20 for output, while GPT-6 Luna cost $0.10 and $0.50, respectively. These prices are based on 1 million tokens for both input and output.
Luna's input unit price is half, and its output unit price has fallen even lower than half. However, it’s important to keep in mind that the comparison is based on the promotional price of the previous model. Actual service costs vary depending on the input-to-output ratio, whether caching is used, and the volume of requests.
| What Sol and Luna Do
OpenAI describes Astra as a top-tier model suited for the most challenging projects. Sol is designed to strike a balance between performance and cost for specialized tasks and coding, while Luna is aimed at high-volume, cost-sensitive repetitive tasks.
The company stated that the two models used training methods similar to those used for Astra. It also noted that they outperformed the previous generation in assessments of reasoning, factual knowledge, coding, computer use, and sorting. Here, “computer use” refers to the model’s ability to navigate apps on a screen and perform multiple steps.
| How Will the Options in the Development Field Change?
For example, you can use Astra for complex designs, Sol for implementation or document reviews, and Luna for simple classification and recurring inquiries. Using different models for different tasks makes it easier to manage costs than sending all requests to a single, high-cost model.
Upon launch, OpenAI announced that it would make both models available in ChatGPT Work, Codex, and the API. Since availability by product and the actual rollout timeline may vary depending on the phased rollout, you should check the screen you're using for details.
| Why Price Alone Isn't the Only Factor to Consider
The price list is the starting point for calculating the total. You have to factor in how many times the model tries to complete the same task, as well as the time it takes to review and revise the responses, to determine the true cost.
What has been finalized this time are the launch dates for Sol and Luna and the API pricing. Since the performance comparison released by OpenAI is based on test settings, we’ll have to wait and see whether the results will translate directly to real-world use cases. The next step for service operators is to compare the success rates and total costs of each model for the same task.
References
Tags
#OpenAI #GPT6 #GPT6Sol #GPT6Luna #GPT6Astra #API Fees #Generative AI #Work Automation #CodingAI #Token Price
