ChatGPT is about to sell a $500 Pro Max package!

ChatGPT is also launching the Pro Max version.

According to widely circulated leaks starting today, OpenAI is preparing to roll out ChatGPT Pro Max with a monthly fee as high as 500 US dollars, which is 2.5 times the price of the existing top-tier plan. Its core selling point is summed up in one sentence: "The fastest Work and Codex."

It is speculated that the $600 mark mentioned here includes a 20% "channel tax"

Two weeks ago, it just suspended new subscriptions to the $200 Pro tier after GPT-6 Astra overwhelmed its clusters;

Behind Astra, a more powerful model codenamed Bel, which is rumored to have produced a proof for the millennium-old Navier–Stokes problem, is queuing up to be released.

With computing power in short supply, the price of the top-tier plan has doubled. As the most advanced productivity tools become increasingly expensive, a realistic question has been brought to the forefront: Will the next generation of truly groundbreaking intelligence become a privilege reserved only for the highest bidders?

What you get for 500 US dollars

On September 24, developer Tibor Blaho and the leak site TestingCatalog successively found traces of Pro Max in ChatGPT's undisclosed front-end configuration: the monthly fee is 500 US dollars, and its main feature is marked as "the fastest Work and Codex".

https://x.com/btibor91/status/2103222629171908937

https://x.com/testingcatalog/status/2103238785278312463

If it is officially launched eventually, it will directly surpass Claude Max (250 US dollars) and Google AI Ultra (about 250 US dollars), becoming the new price ceiling for mainstream personal AI subscriptions.

Different from previous plans, the premium this time does not buy usage quota, but priority of computing power.

Previously, from the 100 US dollar tier to the 200 US dollar tier, users got 5 to 20 times the call limit of the Plus plan;

Pro Max targets time cost — complex tasks of Work and Codex types often run continuously for hours, and for heavy productivity scenarios, low latency is directly equivalent to profits.

It is speculated that its underlying system may be connected to the Cerebras low-latency hardware that OpenAI is already testing.

Astra has drained the computing power reserve

The direct trigger for the price increase is the depletion of available computing power.

GPT-6 Astra, launched on September 3, has Agent capabilities that can directly manipulate systems, and its consumption of inference computing power far exceeds that of previous flagship models.

Just one week after its launch, on September 10, OpenAI urgently suspended new user subscriptions to the 200 US dollar Pro tier.

Tibo, who is in charge of ChatGPT and Codex businesses, publicly stated that this tier "brings the maximum pressure to the system", and OpenAI has not announced a recovery timeline to date.

The commercial logic behind this arrangement is not difficult to understand.

The 200 US dollar tier has become a "heavy all-you-can-eat buffet" exploited by tech enthusiasts and professional institutions, and the platform is actually bearing subsidized losses.

Rather than letting the system be dragged down by this extreme usage, OpenAI takes the opportunity to separate an independent tier with pricing closer to the real inference cost.

The most cutting-edge intelligence is being clearly priced

More powerful capabilities are yet to come, and GPT-6 Astra is only the appetizer.

The official previously revealed that its latest internal model has an overall performance "significantly stronger than Astra" and has completed the proof of the Navier–Stokes equation.

The industry's diversion paths for cutting-edge capabilities are diverging: OpenAI tends to stratify users by commercial price, while Anthropic prefers to set restrictions based on security risks and identity access permissions.

The latest update to Claude's official documentation shows that Opus 5.5 will automatically downgrade high-risk tasks involving offensive cybersecurity, dual-use biology and other related fields to lightweight models for processing, and only users who pass strict identity verification can apply for unblocking.

Moreover, Opus 5.5 will also detect whether users are conducting cutting-edge AI research and development. Once detected, it will degrade model performance, refusing to "make wedding dresses for others".

A monthly fee of 500 US dollars (approximately 3352 RMB) has almost drawn a clear boundary in advance: it is only accessible to enterprises, institutions, or a very small number of individuals with high monetization capabilities.

If core breakthroughs at the Bel level are eventually only opened to users paying 500 US dollars or even more expensive tiers in the future, technological differentiation will likely accelerate into a real gap: people with abundant funds use top-tier inference capabilities to snowball their advantages, while ordinary users are left below the public performance baseline.

But this is also a cold commercial reality.

OpenAI is not a non-profit organization, and the depreciation and power consumption of cutting-edge clusters are unavoidable hard costs.

In the same week, OpenAI also cut the API prices of its lightweight models GPT-6 Sol and Luna by half.

Popularized intelligence is still rapidly deflating and depreciating, and what is really skyrocketing in value is always the extremely scarce pioneer computing power at the top of the pyramid.

There are still unknowns for next week's DevDay

DevDay next Tuesday US time will be the earliest possible occasion for Pro Max to be officially announced, and there are also rumors of a new OpenAI Platform plan starting at 50 US dollars to be released on the same occasion.

There are three points for the outside world to verify then: whether the 500 US dollar pricing will be finally implemented, when the 200 US dollar tier will be unblocked, and at what cost the real next-generation base model Bel will be launched.

After the first DevDay in 2023, due to sudden traffic overload, OpenAI had to close the subscription entrance for the 20 US dollar Plus plan;

Three years later, the new DevDay has not yet opened, but the entrance for the 200 US dollar tier has been closed in advance.

Behind the closed door, a new threshold of 500 US dollars has been quietly written into the code.

References:

https://x.com/testingcatalog/status/2103238785278312463

https://x.com/btibor91/status/2103222629171908937

https://support.claude.com/en/articles/16049681-why-claude-switched-models-in-your-conversation-with-opus-5-or-opus-5-5

Editor: Marco

This article is from the WeChat Official Account "AI_era" (ID: AI_era), author: ASI Revelation, authorized for release by 36Kr.