OpenAI GPT-6 Sol, Luna, and GPT-6.1 Sol bring lower-cost models and stronger prompt caching

OpenAI has introduced GPT-6 Sol and GPT-6 Luna as lower-cost additions to the GPT-6 family. The models are available in ChatGPT Work and Codex for supported paid plans, and in the OpenAI API as gpt-6-sol and gpt-6-luna.

Two lower-cost GPT-6 tiers

OpenAI positions Sol for difficult professional work with more room to iterate and Luna as the faster, more affordable option. In the API, GPT-6 Sol replaces GPT-5.6 Sol at $2 per million input tokens and $10 per million output tokens. GPT-6 Luna is listed at $0.10 per million input tokens and $0.50 per million output tokens.

OpenAI says these prices are 50% below the previous GPT-5.6 Sol and Luna pricing. The API model IDs are gpt-6-sol and gpt-6-luna; availability and pricing should still be checked against the current model documentation before production budgeting.

The models 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, while OpenAI says the models are not yet available in Chat.

Prompt caching is part of the release

OpenAI also describes improved prompt caching for the GPT-6 family. Shared instructions, tool definitions, and carried-forward context can receive up to a 90% discount on cached input tokens when the prefix is reused inside the eligibility window.

The new Prompt Caching Dashboard shows cache hit rates over time and compares cached and uncached input. A diagnostics tool can compare a request with a recent response and identify changes to the model, tools, settings, or input that prevented reuse.

Developers can choose explicit cache breakpoints and change reasoning effort between responses without breaking the cache. Stable tool definitions, schemas, and ordering remain important: changing them can make a shared prefix miss the cache. Prewarming can prepare known instructions, tool definitions, or reference material before the user waits for a response.

Practical migration checks

Treat Sol and Luna as new model IDs rather than assuming a drop-in replacement. Test tool calls, reasoning effort, cache behavior, latency, and output quality on representative agent tasks. For caching, log hit rates and inspect misses before changing prompt structure; a lower token price does not compensate for a poorly reused context.

September 29 follow-up: GPT-6.1 Sol and Ultrafast

OpenAI’s DevDay recap introduced GPT-6.1 Sol, describing it as an upgrade to GPT-6 Sol for agentic coding, computer use, and professional work. The API changelog lists gpt-6.1-sol for the Responses and Chat Completions APIs and says it supports Multi-agent in beta, which lets the model delegate work to subagents in a Responses API request. OpenAI recommends the Responses API for tool calling; teams should verify current pricing and model guidance before changing production defaults.

OpenAI also added Ultrafast service for GPT-6 Astra in the Responses API. Set service_tier: "ultrafast" with gpt-6-astra to reduce the time between generated output tokens. The DevDay recap reports up to 8× faster token generation in Codex and up to 6× in the API; those are OpenAI’s stated maximums, not a latency guarantee for every workload. The recap says Astra Ultrafast is available in the API and in ChatGPT Work and Codex on Pro 500 and Enterprise plans, while GPT-6.1 Sol Ultrafast is coming soon.

Sources

• OpenAI — “Introducing GPT-6 Sol and Luna” (September 22, 2026): https://openai.com/index/introducing-gpt-6-sol-and-luna/
• OpenAI — “Better prompt caching for GPT-6” (September 22, 2026): https://openai.com/index/better-prompt-caching-for-gpt-6/
• OpenAI — “DevDay 2026 Recap” (September 29, 2026): https://openai.com/index/devday-2026-recap
• OpenAI Developers — API changelog: https://developers.openai.com/api/docs/changelog
• OpenAI — Product News and Updates: https://openai.com/news/product-releases/
• OpenAI Developers on X — official developer updates: https://x.com/OpenAIDevs