OpenAI released two new API models, GPT-6 Sol and GPT-6 Luna, on September 22, 2026, cutting per-token pricing in half compared with GPT-5.6’s promotional rates. The bigger cut lands on Luna, the model OpenAI built specifically for high-volume, clearly-scoped tasks — the same profile as most customer-support auto-replies.
What changed
Per OpenAI’s own model documentation, GPT-6 Luna now costs $0.10 per million input tokens and $0.50 per million output tokens, with cached input priced at just $0.01. OpenAI positions it as the fastest, most affordable model in the GPT-6 line, built for “high-volume tasks with a clear goal, like summarizing documents, extracting information, or answering quick questions.” GPT-6 Sol, aimed at complex coding and agentic workflows, prices at $2 input / $10 output per million tokens. OpenAI’s own announcement states both rates sit 50% below GPT-5.6’s promotional pricing. Both models are live now in the API, in ChatGPT on Plus, Pro, Business, Enterprise and Edu plans inside ChatGPT Work, and in Codex; free and Go users get GPT-6 Luna in the desktop app.
Why this matters for Thai sellers and support teams
Plenty of Thai SME support tools already run on OpenAI models under the hood — 1CS, for instance, uses OpenAI’s models for image analysis and knowledge-base search inside its AI Agents. Luna’s use case — high-volume, clearly-scoped replies — is exactly what an AI agent does when it answers a repeat question or looks up an order status. A 50% cut to that model’s price gives vendors more room to lower what they charge per conversation, or to answer more of them for the same budget, and it adds pressure on rival AI providers to match the move.
What to do now
- If your own team calls the OpenAI API directly for auto-replies or ticket summaries, check whether you’re still paying GPT-5.6 rates — GPT-6 Luna is the closer match for that workload.
- If you use a third-party AI support platform, ask whether it has adopted the new models and whether the savings show up on your invoice.
- Re-test answer quality on your own FAQs before trusting a new model fully on customer-facing chats — a cheaper model is only worth it if it stays accurate.
- Keep budgeting per message: Sol and Luna are still priced per token, so cost still scales with how many conversations your AI answers.
Before switching models or vendors, it’s worth checking what to automate before automating everything — cost is only one factor in a good handoff between AI and a person.
