OpenAI's GPT-6 Sol and Luna: Same AI Family, Half the Price

OpenAI just released two new AI models, GPT-6 Sol and GPT-6 Luna. They are cheaper siblings of its flagship GPT-6 Astra — and the headline is the price: running them costs roughly half of what their predecessors cost. Here is what each one is for, and why a price cut matters more than it sounds.

What actually launched

On September 22, 2026, OpenAI expanded the GPT-6 family with two models:

• GPT-6 Sol — the mid-tier workhorse, aimed at the complex work developers and knowledge workers do every day.
• GPT-6 Luna — the budget option, built for high-volume, routine jobs.

Both improve on their GPT-5.6 predecessors in OpenAI's benchmarks, but both sit below the flagship GPT-6 Astra, which remains the top choice for the hardest problems.

Where you can use them: through the OpenAI API as gpt-6-sol and gpt-6-luna, and rolling out in ChatGPT Work and Codex for Plus, Pro, Business, and Enterprise customers. People on the Free and Go plans can reach Luna through the desktop app. They are not in the standard ChatGPT chat window yet.

The prices, in plain terms

AI models charge by the "token" — roughly, a small chunk of text. Apps that use AI can burn through millions of tokens, so the per-token price decides real bills. Here is the new price list:

• GPT-6 Sol · Input (per 1M tokens): $2.00 · Output (per 1M tokens): $10.00 · Change vs. predecessor: 50% cheaper than GPT-5.6 Sol
• GPT-6 Luna · Input (per 1M tokens): $0.10 · Output (per 1M tokens): $0.50 · Change vs. predecessor: 50% cheaper input, ~58% cheaper output than GPT-5.6 Luna
• GPT-6 Astra (flagship) · Input (per 1M tokens): $10.00 · Output (per 1M tokens): $50.00 · Change vs. predecessor: Unchanged

Three things worth knowing:

1. These are permanent prices, not a sale. An OpenAI spokesperson confirmed the rates have no promotional expiration — companies can plan around them.
2. Sol is priced exactly at Claude Sonnet 5's level ($2/$10) — a direct challenge to Anthropic's most popular model.
3. Both models handle up to 1.05 million tokens of context — that is roughly a whole bookshelf of text in one go.

Which one is for what

Think of it like vehicles: you would not use a race car to deliver groceries.

• Sol is the work truck. Best for repeated, meaty work: building software features, reviewing code, debugging, analyzing data. OpenAI reports it scoring 33.2% on its AutomationBench agent test at about $0.27 per task.
• Luna is the delivery van. Best for simple jobs done at huge scale: summarizing documents, extracting data, answering straightforward questions. At these prices, routing routine agent work to Luna instead of a big reasoning model slashes costs dramatically.
• Astra is the race car. Reserved for the hardest, multi-part problems — tough science, math, and complex projects mixing text, images, and more.

OpenAI's pitch is that companies should stop asking "which model is best?" and start asking "which model is cheapest for this specific job?"

Why this matters to you (even if you never touch an API)

You may never pay per token yourself. But cheaper models change the apps you use:

• AI features get cheaper to offer, so more tools can include them without raising prices.
• AI agents that do multi-step work — searching, comparing, drafting, checking — become affordable, because every step of an agent's work burns tokens.
• Caching got cheaper too: OpenAI now discounts reused prompt content by 90%, which matters because agents re-send the same instructions over and over.

What to be careful about

• Vendor benchmarks deserve skepticism. OpenAI's charts compare Sol against Anthropic's older Opus 5 — but Anthropic released Opus 5.5 the very same day. There is no public same-test comparison of Sol vs. Opus 5.5 yet, so treat "cost per task" claims as directional, not final.
• Cheaper does not mean equal. Luna costs 95% less than Sol per input token, but it is built for simpler jobs. Price and capability are different things.
• You cannot try them in normal ChatGPT yet. For now, these are mainly developer and business tools.

Explain it to a friend: "OpenAI released two cheaper AI models. Sol is the affordable all-rounder for coding and office work; Luna is the super-cheap one for simple, repetitive jobs done at scale. Both cost about half of what the older versions did — and the prices are permanent."

Try this: If an AI tool you use lets you pick a model, try the cheaper tier for routine tasks first. You often will not notice the difference — and your credits will last much longer.