OpenAI just made its GPT -6 models much cheaper
OpenAI has cut prices for GPT -6 models Sol and Luna, making advanced AI cheaper for developers, startups and businesses building at scale.
OpenAI has introduced GPT-6 Sol and GPT-6 Luna, two new models designed to bring its latest AI capabilities to more everyday business and developer use cases at a lower cost.
The models carry forward several improvements from GPT-6 Astra, OpenAI’s most powerful model, but focus more heavily on speed, affordability and wider access.
The biggest change is the price. OpenAI has cut API pricing for Sol and Luna by 50% compared with GPT-5.6 promotional pricing, potentially giving developers more room to build and scale AI products without costs rising as quickly.
Sol and Luna bring down the API bill
For GPT-6 Sol, input pricing has fallen from $4 to $2 per 1 million tokens, while output pricing has dropped from $20 to $10. Tokens are the small units of text that AI models process when reading prompts and generating responses.
For companies running AI products at scale, even small changes in the cost per token can add up quickly.
GPT-6 Luna goes further on price. Its input cost is now $0.10 per 1 million tokens, down from $0.20, while output pricing has fallen from $1.20 to $0.50.
That makes Luna the cheaper option for high-volume workloads, while Sol is aimed at tasks where businesses still need stronger reasoning and output quality.
For Indian startups and software teams building AI assistants, coding tools, customer support bots and workflow automation, lower API costs could make experimentation and scaling easier.
Two models, different jobs
OpenAI says GPT-6 Astra remains its highest-quality model. Sol and Luna sit below it, with both designed around cost efficiency across areas including professional work, factuality, coding, computer use and alignment.
Sol is targeted at more demanding workloads. OpenAI says it performs strongly on business workflows, coding and computer-use benchmarks while costing less per task than several competing models in its comparisons.
Luna takes the lighter route. Its lower price makes it more suitable for applications where speed and cost matter alongside performance, including routine support, summarisation, content assistance, simple coding and internal productivity tools.
Caching could make long AI workflows cheaper
OpenAI is also changing how developers pay for repeated context. With prompt caching, repeated parts of a prompt or long conversation can be reused instead of being processed from scratch each time.
OpenAI says cached input-token reads can receive discounts of 90%. That could matter particularly for AI agents and long-running workflows that repeatedly use the same instructions, files or background context.
AI pricing is becoming part of the product
GPT-6 Sol and GPT-6 Luna are available in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu users. Free and Go users can access GPT-6 Luna in the desktop app. The models are also available through the API.
The price cuts point to a broader change in the AI market. As models become more capable, businesses are increasingly looking beyond raw intelligence.