GPT-6 Sol and Luna: 7 Amazing Facts About the New Models

On September 22, 2026, OpenAI quietly reshaped the economics of AI work. Alongside its flagship GPT-6 Astra, the company released GPT-6 Sol and Luna — two new models built for everyday professional tasks at a fraction of the flagship’s cost. If you use AI to write, code, summarize, or run automated workflows, this launch changes your math in a serious way.

Below is everything confirmed about GPT-6 Sol and Luna so far: what they are, what they cost, how they differ, and what they mean for creators and businesses. All facts come from OpenAI’s launch announcement and independent reporting.

What Are GPT-6 Sol and Luna?

GPT-6 Sol and Luna sit one tier below the flagship GPT-6 Astra in OpenAI’s new model family. OpenAI says the pair inherits much of Astra’s progress in reasoning, factual reliability, coding, computer use, and alignment — while being tuned for speed and cost efficiency.

The company’s framing is straightforward: instead of reserving frontier-level intelligence for only the most demanding projects, GPT-6 Sol and Luna spread that capability across more tasks and budgets. They are designed for the recurring, everyday work that eats up API bills, not for once-in-a-while showcase queries.

According to the launch announcement, the two models are aimed at different workloads. GPT-6 Sol targets recurring complex work: coding, debugging, building features, data analysis, and agentic workflows. GPT-6 Luna is aimed at high-volume, well-defined jobs: document summarization, information extraction, and short-answer questions.

GPT-6 Sol Pricing

OpenAI’s posted standard API rates for GPT-6 Sol are $2 per million input tokens and $10 per million output tokens. The company attributes the reductions to more efficient caching and inference, and says it is passing those savings to customers.

GPT-6 Luna Pricing

GPT-6 Luna is priced at just $0.10 per million input tokens and $0.50 per million output tokens. At that level, summarizing thousands of documents or running extraction over a huge dataset costs less than a cup of coffee.

7 Amazing Facts About GPT-6 Sol and Luna

1. They Launched Less Than Three Weeks After GPT-6 Astra

GPT-6 Astra arrived in early September 2026 as the new flagship. GPT-6 Sol and Luna followed on September 22 — a fast follow-up that shows how quickly OpenAI is expanding the family. If you are tracking this month’s launches, our September 2026 AI news roundup covers the full wave, including Astra, Claude Opus 5.5, and Grok 4.7.

2. API Prices Are 50% Lower Than the GPT-5.6 Generation

OpenAI says API rates for both models are 50 percent lower than the promotional prices of their GPT-5.6 counterparts. That kind of cut does not happen often at the frontier, and it comes from engineering improvements in caching and inference rather than reduced capability.

3. Sol Makes Roughly Half as Many Factual Errors as Its Predecessor

According to OpenAI’s internal evaluations drawn from real user conversations, GPT-6 Sol produces roughly half as many factual errors as GPT-5.6 Sol. For creators and teams who rely on AI-drafted research and scripts, reliability gains like that matter more than raw benchmark scores.

4. Sol Can Match Pricier Rival Models on Coding Tasks

OpenAI highlights gains on coding and computer-use benchmarks, claiming Sol can match or exceed some higher-priced rival models at a fraction of the cost per task. Early user reports, including from the automation platform Zapier, describe stronger results on automation benchmarks at lower cost, with Sol showing better attention to multi-step tasks than the prior generation.

5. Luna Is Built for Volume, Not Depth

The interesting design decision in GPT-6 Sol and Luna is the split: Sol handles recurring complex work, while Luna handles high-volume, well-defined jobs. That split mirrors how businesses actually spend on AI — a few expensive thinking tasks and a huge number of cheap routine ones. Luna’s pricing turns those routine jobs into a rounding error.

6. They Are Available Beyond the Raw API

The launch announcement distinguishes access channels: the models are rolling out through the hosted API, ChatGPT Work, and Codex. That matters because many creators will meet GPT-6 Sol and Luna inside the tools they already use, without touching an API key.

7. They Anchor a Month of Cheaper Hosted Intelligence

Zoom out and the pattern is clear. September 2026 also brought Grok 4.7 at $2/$6 API pricing, expanded Gemini 3.8 models, and more. The industry-wide trend is hosted intelligence getting cheaper while downloadable weights still carry very different permissions. For anyone building AI into a content workflow, the cost floor keeps dropping.

What GPT-6 Sol and Luna Mean for Creators

For content creators, the headline is not a leaderboard number — it is the cost of a finished, checked task. As one industry analyst put it, the practical test of a cheaper model is the price of one edited article, one working feature, or one cleaned-up spreadsheet. A cheap answer that needs three rounds of repairs is less cheap than it looks.

Here is what GPT-6 Sol and Luna realistically unlock for creators:

• Cheaper script iteration. Drafting and revising video scripts with Sol costs far less per round, so you can afford more passes.
• Batch content operations. Luna’s pricing makes it practical to summarize research, extract quotes, and generate video descriptions in bulk.
• Agentic repurposing pipelines. The kind of multi-step workflows that turn one video into a blog post, a newsletter, and ten social posts get meaningfully cheaper on Sol.

Pair the new models with a disciplined system and the savings compound. Our guide to turning AI prompts into daily content shows how batching prompts and scheduling ahead turns cheap AI output into a consistent publishing machine. And if you want to keep your tooling spend at zero while you experiment, these free AI tools for content creators cover editing, design, voiceovers, and images without a credit card.

How GPT-6 Sol and Luna Compare to the Alternatives

The honest comparison is about price per finished task, not price per token. On that basis, Sol’s combination of lower factual-error rates and stronger multi-step attention is its real edge: fewer repairs means the headline price is closer to the true price.

For high-volume, well-defined jobs, Luna’s $0.10/$0.50 rate is hard to argue with. The workloads it targets — summarization, extraction, short answers — are exactly the jobs where a flagship model’s extra reasoning is wasted money.

None of this makes the flagship irrelevant. GPT-6 Astra still leads for the hardest reasoning and computer-use tasks. But the practical lesson of the GPT-6 Sol and Luna launch is routing: match the job to the model, and keep the expensive thinking for the jobs that need it.

GPT-6 Sol and Luna: FAQ

When were GPT-6 Sol and Luna released?

OpenAI released GPT-6 Sol and Luna on September 22, 2026, about three weeks after the flagship GPT-6 Astra arrived in early September.

How much do GPT-6 Sol and Luna cost?

GPT-6 Sol costs $2 per million input tokens and $10 per million output tokens. GPT-6 Luna costs $0.10 per million input tokens and $0.50 per million output tokens — both 50% lower than the promotional prices of their GPT-5.6 counterparts.

Should I choose Sol or Luna?

Choose Sol for recurring complex work like coding, debugging, data analysis, and agentic workflows. Choose Luna for high-volume, well-defined jobs like document summarization, information extraction, and short-answer questions.

Where can I access GPT-6 Sol and Luna?

According to OpenAI’s launch announcement, the models are available through the hosted API as well as ChatGPT Work and Codex, with different plan and rollout details for each channel. Check OpenAI for the current availability in your region.

Cheaper, more reliable models do not replace good prompting and good systems — they amplify them. If you are ready to put that to work, browse our AI News section for weekly plain-language coverage of every launch that actually matters.