Anthropic has extended the availability of Claude Fable 5. This most powerful AI model was originally scheduled to be removed from Pro, Max, and Team subscriptions yesterday, but that deadline has now been pushed back to July 12. Meanwhile, OpenAI is the verge of making GPT-5.6 widely available after this LLM release was also blocked by the U.S. government.
Fable 5 will be available to all paying Claude users through July 12. Up to 50 percent of the total weekly token allowance may be used to run this model, although 5-hourly limits prevent it from being used up in a very short period of time. Anyone who exceeds the limit will be charged the substantial usage-based fees that also apply to the API.
OpenAI prepares a response(?)
Fable 5 is impressive, just as it was when it became available in early June. However, that promotional period, originally intended to run through June 22, was cut short. The U.S. government received a tip, reportedly via Amazon CEO Andy Jassy, that Fable 5 had a jailbreak. Anthropic described the alleged jailbreak as “limited,” but the U.S. imposed a sudden export ban late on Friday afternoon, June 12.
The saga had an immediate impact on OpenAI. While GPT-5.6 is available in preview for 20 approved organizations, widespread availability has not yet been achieved, much to the ChatGPT creator’s dismay. However, Axios reports that GPT-5.6 will be released to the market on July 9 following Fable’s lead. The released benchmarks are limited but suggest that the various variants (Sol, Terra, Luna) are competitive with all LLMs except Fable, and that Sol, in so-called Ultra mode, might well be competitive at that highest level.
The price war seems to be over for now
Fable 5 will therefore disappear behind a usage-based paywall at a later date than expected. But that doesn’t mean it will stay that way: Anthropic has repeatedly indicated that it will restore availability based on its own computing capacity. With an estimated size of 2 to 10 trillion parameters, Fable 5 and Mythos 5 place a very heavy workload on Anthropic’s infrastructure; due to its strong B2B focus, the company finds it harder to explain downtime to customers than it would with a consumer chatbot.
In any case, while prices for smaller models may be falling, and capable open-source alternatives from China, such as GLM-5.2 and Tencent’s recently released Hy3, are relatively very cheap per token, that doesn’t tell the whole story. The very largest LLMs enable new use cases and handle a larger share of the work (particularly in coding) without requiring user “babysitting.” That means a price war at the highest quality level isn’t very likely for the time being. Many have pinned their hopes on OpenAI, which, with its own Codex/ChatGPT, offers a much more generous monthly usage limit than Anthropic does with Claude. Still, it seems Fable 5-quality isn’t something we should expect from GPT-5.6 just yet.
Read also: Claude Fable 5 is Mythos for the masses