OpenAI's guide to GPT-6 shows the cheapest model costs one hundredth of the top one

OpenAI published a practical guide to its GPT-6 family. It lists three models with their prices per million tokens of input and output: Astra at $10 and $50, Sol at $2 and $10 (near-Astra quality for a fifth of the price), and Luna at $0.10 and $0.50 for fast everyday work. The advice is to match the model to the job, reuse repeated instructions through caching, which can cut input costs by up to 95 percent, and say clearly what done looks like.

• If a tool you pay for lets you choose the model, try the middle one first for everyday jobs such as drafting replies and summarising calls, and move up only for work where it clearly falls short.
• Ask any developer or agency building you an AI feature which of the three models it uses and why, and whether repeated instructions are cached to keep the bill down.

Source: OpenAI. This is our short summary of their reporting. The paragraphs marked "Our read" and the things to do are SeedFoundry's opinion. The summary is drafted with AI tools. Image: OpenAI, from its own announcement. How we source and credit