GPT5.6 was drastically scaled back, is Fable 5 set to return as a severely underpowered version?
GPT-5.6 Sol was split and held down. Fable 5 was globally banned for 72 hours and then returned in shackles. The most powerful models of Anthropic and OpenAI were both "brain-cut."
Yesterday, OpenAI's most powerful series, GPT-5.6, suddenly went live.
However, GPT-5.6 was split into three forms:
- GPT-5.6 Sol: The next-generation cutting-edge model.
- GPT-5.6 Terra: A balanced model suitable for daily efficient work.
- GPT-5.6 Luna: A fast and cost-effective model for high-throughput tasks.
However, the release right of the most powerful GPT-5.6 Sol is no longer in the hands of Altman of OpenAI, but in the "approved list."
Before OpenAI backed down, Anthropic had just experienced a "ban crisis": Except for some partner institutions, Fable 5 was globally banned.
In the era of algorithm rationing, GPT 5.6 was brutally brain-cut
Previously, products were recalled when there were problems. This time, the products were stopped at the door before they even left.
According to the source, OpenAI was required to limit the release scope of GPT-5.6.
The reason is to hold down the most powerful models before the AI security framework across the United States is implemented.
It is revealed that this intervention is because GPT-5.6 has the ability "similar to Mythos," rather than a sudden and tougher intervention measure.
After careful consideration, OpenAI decided to limit the access rights of the latest AI. Only a limited number of users were able to preview GPT-5.6.
The vast majority of ChatGPT users still need to wait for a few weeks for GPT-5.6 to be gradually lifted from the ban.
Mythos 5 is terrifying! Will Fable 5 return as a weak version?
Before OpenAI backed down, another giant, Anthropic, had just experienced a crisis.
Two weeks ago, Anthropic released the mythical AI model Fable 5.
However, this model only survived on the shelves for 72 hours before being violently taken off the market.
It's because of fear. The network attack ability of Fable 5 is terrifying.
Andrew Garbarino, a member of the U.S. House of Representatives, said so.
In a closed-door demonstration, Anthropic Mythos showed an amazing threat to the software security field. For example, it could even empty a private bank account.
Andrew Garbarino said:
Anthropic instructed the model to find the loopholes in the banking system and empty the accounts, and then it carried out the task.
Subsequently, Mythos "could also find this loophole and fix it."
This is an unprecedented "pull the plug" operation in human history.
Until recently, Mythos 5 was allowed to be opened to a very small number of "trusted partners" in a "shackled" state.
According to the latest report: The "castrated version" of Mythos 5, Fable 5, is expected to return and go live next week.
However, Fable 5 may be severely castrated to ensure that the security guardrails are stronger.
When Fable 5 was first released, some users found that as long as they said the word "cancer," they would be blocked by Claude.
If the guardrails are further strengthened, the user experience may be a nightmare.
Mikhail Parakhin, the CTO of Shopify, has deeply tested Fable 5 and GPT-5.6 Max.
He believes that Anthropic no longer has an absolute advantage, and GPT-5.6 has caught up with Claude Mythos.
GPT 5.6 is significantly better than Opus 4.8 in all aspects (it is also slightly faster, but it depends on the load).
Compared with Fable 5, it is significantly inferior in coding, but performs better in agent workloads.
This is a rare real comparison between Fable 5 and GPT-5.6, but it is enough to show that Anthropic cannot rest easy.
If Fable 5 returns only in a castrated version, it will undoubtedly leave more opportunities for competitors to catch up.
Did you buy a genius or a puppet?
What worries developers the most is not only "not being able to buy" but also "getting the wrong product."
In the AI circle, a word is spreading like a virus: "Lobotomized."
In order to pass those complicated security reviews, AI giants have to install layers of strengthened "guardrails" on the models.
However, intelligence often arises from chaos and freedom.
When you forcibly castrate a model's exploration ability in the underlying code, encryption algorithms, and even deep logic, what you get may no longer be an all-knowing "Sol," but a mediocre "Luna" with its claws cut off and meekly submissive.
You want the "fire of Prometheus," but the United States only allows you to have an exported "lighter."
Developers generally fear that the models will have their guardrails significantly strengthened, or even be "lobotomized," with their abilities castrated and severely reduced in intelligence.
Even worse, it may be only available in the United States or require strict identity verification, rather than the "widely available and payable" service promised at the beginning.
There is also the issue of money.
Anthropic initially provided Fable 5 for free in multiple paid Claude subscriptions until June 22, giving users a short window to experience its power - and then, the access disappeared.
Now, no one can say for sure: Can the subscribed users still get the "free and unrestricted use" promised at the beginning? Or, when it returns, will it be locked behind the doors of extra fees or identity verification?
Maybe soon, we will miss that spring in 2023. At that time, as long as you had an email, you could command the "god-like AI."
Reference materials:
https://punchbowl.news/article/tech/garbarino-mythos/
https://www.axios.com/2026/06/27/anthropic-fable-5-return-soon
https://www.wsj.com/tech/ai/openai-limits-access-to-new-model-citing-government-security-concerns-66420050
This article is from the WeChat official account “New Intelligence Yuan”, author: ASI Apocalypse. It is published by 36Kr with authorization.