GPT-6 Rollout: Pricing, Sol vs Luna, and the New $500 Plan Explained
GPT-6 Sol and Luna are live in ChatGPT, with a new $500/month tier, halved usage caps on $200 plans, and a new Decisions API.
What is GPT-6 and who gets access to it?
OpenAI has rolled out GPT-6 across ChatGPT, splitting access by plan. Paid subscribers get GPT-6 Sol, the full version of the new model, while free users get GPT-6 Luna, a lighter variant built for lower compute cost. OpenAI also shipped GPT-6.1 Soul, which it says comes close to the performance of its flagship model, GPT-6 Astra, but runs cheaper. Alongside the model launch, OpenAI changed pricing: a new $500 a month tier appeared, and usage limits on the existing $200 plans were reportedly cut roughly in half.
TL;DR
- GPT-6 Sol is the full model available to paying ChatGPT users, while GPT-6 Luna is a scaled-down version for free-tier users.
- OpenAI introduced intelligent UI, where ChatGPT can respond with interactive elements like charts, calculators, and buttons instead of plain text blocks.
- GPT-6.1 Soul reportedly performs close to the top-tier GPT-6 Astra model at a lower cost, giving users a middle option between Sol and the most powerful tier.
- OpenAI added a $500 a month plan while reportedly halving the usage allowance on the $200 plans, a change that frustrated heavy users.
- The new Decisions API is built for fast, cheap, simple decisions (reportedly around a tenth of a second), aimed at developers building things like responsive voice assistants.
- The rollout lines up with OpenAI reportedly closing a bridge funding round of up to $30 billion at a valuation near $1.25 trillion, showing how much financial pressure sits behind these product changes.
How does GPT-6 differ from previous ChatGPT models?
The headline feature is intelligent UI. Instead of returning a wall of text, ChatGPT can now generate interactive components tailored to the request. Ask it to compare mortgage options, for example, and it can produce a working calculator rather than a paragraph explaining the math. This shifts ChatGPT from a pure text generator toward something closer to a dynamic app interface that adapts its output format to the task.
The tiered rollout (Sol for paying users, Luna for free users) also signals a more deliberate split in capability by plan than OpenAI has used in past launches. GPT-6.1 Soul adds a third option, positioned as a near-flagship model at a lower cost than the full GPT-6 Astra tier, giving users and developers more granularity in choosing a tradeoff between capability and price.
Why did OpenAI raise prices and cut usage limits?
The $500 a month plan and the reported halving of usage caps on $200 plans mark a clear shift for power users. If you were getting heavy daily use out of a $200 subscription, that same spend now reportedly buys roughly half the usage it did before, pushing anyone who needs more back toward the new, pricier tier.
This fits a broader pattern already visible in the rest of OpenAI’s announcements: make the base model smarter and more interactive for everyone, make it faster and cheaper for developers building at scale (via the Decisions API), and charge the heaviest individual users significantly more. It’s a pricing structure built around segmenting demand rather than a flat one-size-fits-all subscription.
The timing also lines up with reports that OpenAI is close to closing a bridge funding round of up to $30 billion, led by investors from the UAE and investment firm Coatue, at a valuation of around $1.25 trillion before the new money lands. That’s lower than the $1.4 trillion figure that had circulated earlier, but it’s still a valuation most public companies never reach. A separate compute deal in the UAE was also reportedly under negotiation. Running models at this scale costs enormous amounts of compute, and pricing changes like these are one of the clearest ways that cost shows up for end users.
What is the Decisions API and why does it matter?
The Decisions API is a developer-facing tool designed to make simple decisions very quickly and cheaply, reportedly in around a tenth of a second. It’s not meant for complex reasoning tasks. It’s meant for the kind of lightweight, fast calls that make AI feel instant inside an application, like a voice assistant that responds immediately instead of pausing awkwardly while it “thinks.”
Auditable. Scalable.
This matters because a lot of AI product friction right now comes from latency. A chatbot that takes a few seconds to answer a yes/no question feels broken in a voice interface or a live app. By offering a stripped-down, fast, cheap endpoint specifically for quick decisions, OpenAI is targeting the plumbing layer of AI products rather than the flashy model-capability layer. It’s a sign that OpenAI sees a market not just in being the smartest model, but in being the fastest and cheapest option for the huge volume of small decisions that real apps need to make constantly.
Is the new pricing worth it for heavy users?
Whether the $500 plan is worth it depends entirely on how much you were running into limits on the $200 tier. If the reported halving of usage caps is accurate, users who were already near their limits on the $200 plan will likely hit a wall much faster now, making the jump to $500 a practical necessity rather than a luxury upgrade.
For casual or moderate users, this change probably doesn’t register at all. Free users get GPT-6 Luna, which still represents a capability upgrade over whatever model preceded it, just scaled down. The real pressure point is the middle and top of the user base: people and small teams who were leaning on the $200 plan for serious daily work. For them, the practical choice now is to either absorb the higher cost, change workflows to use less per session, or look at competing models from Anthropic or Google as a cheaper alternative for some tasks.
It’s also worth watching whether this becomes a broader industry pattern. Anthropic’s recent Claude Sonnet 5.5 release, which the company says is around 30% faster and 30% cheaper than its predecessor, points the opposite direction, toward driving costs down for the core workhorse model. If rivals keep undercutting OpenAI on price for comparable capability, that puts pressure on how sustainable OpenAI’s higher-tier pricing is over time.
What does this rollout say about where ChatGPT is headed?
Put together, the GPT-6 launch shows OpenAI pulling in two directions at once. On one side, it’s making the core product more capable and more useful for free users (Luna, intelligent UI) and cheaper and faster for developers building on top of it (Decisions API, GPT-6.1 Soul). On the other, it’s extracting more revenue from its most intensive users through the new $500 tier and tighter caps on the $200 plan.
That’s consistent with a company facing enormous financial pressure to justify a valuation above a trillion dollars while continuing to pour money into compute and research. Shipping a smarter model isn’t optional when you’re raising tens of billions of dollars and competing directly with Anthropic, which is reportedly preparing for an IPO targeting a valuation around $2 trillion. Expect usage-based pricing experiments and tiered access to keep evolving as these companies look for ways to make the economics work without slowing down model releases.
Frequently Asked Questions
What’s the difference between GPT-6 Sol and GPT-6 Luna?
GPT-6 Sol is the full version of GPT-6 available to paying ChatGPT subscribers. GPT-6 Luna is a lighter, less resource-intensive version given to free-tier users.
How much does the new ChatGPT $500 plan cost and who is it for?
The new tier costs $500 a month and is aimed at heavy users who need usage levels beyond what the $200 plan now offers, especially after that plan’s usage allowance was reportedly cut roughly in half.
What is GPT-6.1 Soul?
GPT-6.1 Soul is a model OpenAI released that reportedly performs close to its top-tier model, GPT-6 Astra, but at a lower cost, giving users a mid-tier option between Sol and the most powerful (and presumably most expensive) model.
What does the Decisions API actually do?
It’s a developer tool for making fast, simple decisions very cheaply, reportedly in about a tenth of a second, intended for use cases like voice assistants or apps that need near-instant responses rather than deep reasoning.
Did usage limits actually get cut in half on existing plans?
According to reporting at the time of the rollout, usage allowances on the $200 plans were reduced by roughly half alongside the introduction of GPT-6 and the new $500 tier.
