OpenAI is reportedly preparing a $500 per month subscription tier that would push the ceiling on what consumers pay for a personal AI assistant to a figure most would consider absurd. The new "Pro Max" plan would sit above the existing $200 Pro tier, and the announcement has landed at a moment when the company's own infrastructure is straining under demand.

How OpenAI arrived at a five-tier pricing ladder

The current ChatGPT pricing structure has expanded rapidly over the past two years. OpenAI launched its $200 per month Pro plan on December 5, 2024, marketing it as unlimited access to the company's most advanced reasoning models for "researchers, engineers, and other individuals who use research-grade intelligence daily." Six months later, on April 9, 2026, the company added a $100 Pro tier offering five times the usage limits of the $20 Plus plan. That move was explicitly framed as a competitive response to Anthropic, whose Claude Max 5x plan had launched the same month at the identical $100 price point.

The result is now five personal tiers: Free at $0, Go at $8, Plus at $20, Pro at $100 with 5x usage limits, and Pro at $200 with 20x limits. Both Pro tiers share the same core models and features. The only variable is how much you can use them before hitting a ceiling. A $500 tier would add a sixth level, costing 25 times the base Plus plan and 2.5 times the current top-end Pro subscription.

The infrastructure problem behind the price hikes

On September 10, 2026, OpenAI paused new subscriptions and upgrades to its $200 Pro tier. The reason was the overwhelming demand for GPT-6 Astra, the company's most recent model released on September 3. Existing subscribers were unaffected, but anyone attempting to sign up or upgrade to the top tier found the door closed. New sign-ups for lower tiers continued without interruption.

Research firm SemiAnalysis provided a striking data point in June: a $200 per month subscription to ChatGPT Pro yields roughly $14,000 worth of token volume, while the equivalent Claude Max plan delivers about $8,000. The gap between what users get and what they pay is enormous, but so is the compute cost of delivering it. OpenAI is not raising prices out of greed alone. The company is managing demand for models that are consuming compute at a pace its infrastructure was not built to sustain.

What the HN thread reveals about user sentiment

The Hacker News discussion around the $500 figure captures a frustration that goes beyond the sticker shock. User "preommr," posting on the thread, pointed to a deeper problem with how OpenAI structures its subscriptions: the opacity. Model labels shift, usage multipliers get rebranded, and subscribers who paid for 5x or 20x access in good faith watch those tiers get reorganized without clear communication.

The concern that existing feature sets could be redistributed into a new, more expensive tier is not irrational. OpenAI has a pattern of adjusting what "unlimited" actually means, and the recent rebalancing of Codex usage in the Plus plan in April, which shifted from long single-day sessions to more distributed weekly sessions, demonstrated how quickly usage terms can change after launch.

What a $500 tier means for developers and teams

At $500 per month, this plan is not aimed at individual developers or small teams. It is aimed at enterprises and power users for whom AI is a production dependency rather than a convenience tool. The comparison to Anthropic's Claude Max, which tops out at $100 per month, and Google's Gemini Advanced at $20 per month, puts OpenAI in a unique position: it is the only major lab considering a consumer subscription that breaches the $200 barrier.

For the developer standing at the $20 or $100 tier, the practical implication is that the ceiling on what OpenAI considers "heavy use" keeps rising. The $100 tier launched as an accessible upgrade path for heavier daily Codex use. If the $500 tier materializes, the company is signaling that even the $100 tier may not be sufficient for the workloads its most demanding customers are running. The coding assistant market, where both Codex and Anthropic's Claude Code have seen explosive growth, is the primary battleground, and pricing is becoming a proxy for compute access.

The uncomfortable question

There is a simpler question lurking behind all of this: at what point does the subscription model stop making sense for the end user? With token volume on the $200 plan already representing thousands of dollars in API value, and OpenAI reportedly exploring a tier five times that price, the math increasingly favors API pricing for anyone who actually understands their own usage patterns. The subscription model works because it abstracts away the complexity. But as OpenAI adds more tiers, changes usage definitions, and pauses sign-ups, that abstraction is breaking down.

The $500 plan has not been officially confirmed. OpenAI has not published pricing or feature details for it. But the fact that the number has entered the public conversation, and that users are already organizing around their frustration with it, suggests the company may be closer to unveiling it than it realizes.