Announcing GPT-6 Sol and Luna: 50% API price cut, rolling out to Plus, Pro and Codex today

OpenAI just dropped GPT-6 Sol and GPT-6 Luna. The two new models slot into the GPT-6 family below Astra, inheriting most of its strengths but at faster speeds and lower cost. Pricing takes a clear cut: API calls for Sol and Luna are 50% cheaper than the GPT-5.6 promotional rate. That’s a meaningful drop if you’re running workloads at scale.

Both models roll out today across ChatGPT Work, Codex, and the API. Paid tiers—Plus, Pro, Business, Enterprise, and Edu—get immediate access. Free and Go tier users can try Luna from the desktop app.

What’s different between them? Sol leans into the higher‑end capability end of the spectrum, while Luna is tuned for efficiency and lower‑cost workloads. The trade-off is familiar: more intelligence versus more iterations per dollar.

OpenAI also flagged efficiency gains on the infrastructure side. Caching and inference have been re‑worked, and those savings get passed straight to the user. If you’ve been priced out of running GPT‑5.6 at volume, Sol and Luna may finally make it feasible.

Developers already on the API can start switching today. The model names in the endpoint are gpt‑6‑sol and gpt‑6‑luna. Rate limits are higher than GPT‑5.6, and the cost per token is explicitly lower. OpenAI says the goal is to give “more flexibility and room to iterate” without the bill shock.

For ChatGPT users, Sol and Luna show up in the model selector alongside Astra. No extra setup is needed—just pick the one that matches the task. Codex users get code‑generation improvements out of the box, with the model handling more context before hitting a wall.

OpenAI hasn’t published a full benchmark suite, but the pattern is consistent: newer model, lower price, higher caps. If you’ve been waiting for the price to drop before moving a production workload off GPT‑4‑turbo, this is the sign‑off.

Key details at a glance

• Pricing: 50% lower than GPT‑5.6 promotional pricing
• Availability: API, ChatGPT Work, Codex; Luna also in desktop app for free/Go users
• Release date: today
• Target: Plus, Pro, Business, Enterprise, Edu paid tiers (free users get Luna only)
• Focus: higher usage limits + lower cost = more room to iterate

If you’ve been on the fence about scaling GPT‑6 workloads, the numbers make a strong case. Lower unit cost + higher caps = a lower barrier to entry for anything beyond prototyping.