GPT-6 Luna’s API Rates Offer a Smaller Budget—If It Can Finish Your Task
News / AI tool budgets
A fixed-token example shows the price difference, while real savings depend on quality, retries and total usage.
Source information checked on October 3, 2026 at 07:56 (UTC+8). This brief uses official documentation and stated editorial calculations; the site has not tested the product or verified an individual account’s eligibility or bill.
OpenAI released GPT-6 Sol and Luna in the API on September 22. For Standard processing with prompts up to 272K input tokens, the changelog lists Luna at $0.10 per million input tokens and $0.50 per million output tokens; Sol is $2 and $10 respectively. Official API changelog
For an illustrative monthly workload of one million uncached input tokens and 100,000 output tokens, those rates produce token charges of $0.15 for Luna or $3 for Sol. The example excludes cache writes, tool charges and other processing conditions. It assumes identical usage, not identical results.
For freelancers building classification or extraction workflows, test whether the smaller model meets your acceptance standard before switching. Count retries and human corrections, then compare cost per completed job.
Luna’s model documentation specifies different rates for longer prompts and other processing options. Official model documentation
These are API charges, not ChatGPT subscription prices. Use our automation cost guide to add the platform and maintenance costs around the model.
Limits and unverified details
The USD example uses Standard processing, each prompt up to 272K input tokens, one million uncached input tokens and 100,000 output tokens. It excludes cache writes, tools and other processing conditions. Equal usage does not mean equal results or measured 95% workflow savings; regional processing and account conditions are unverified.