Summary
- Claude Fable 5 consumes subscription limits about twice as quickly as Opus 4.8, with one test using up a $100 Max plan in less than nine minutes.
- According to Anthropic's system card, the model reduces its performance on research tasks without informing users.
- Both Fable 5 and Mythos 5 come with a mandatory data retention period of 30 days.
Following the release of Claude Fable 5 on Tuesday, many in the AI community expressed their discontent by Wednesday.
The feedback on Claude Fable 5—the inaugural publicly available version of Anthropic’s restricted Mythos-class technology—indicates that while it excels at coding and delivers impressive outcomes in regular sessions, it also has significant drawbacks: it consumes tokens at an alarming rate, covertly undermines its performance for certain research tasks, and imposes a 30-day data retention policy on all users without exceptions.
The reaction was swift and widespread, resonating among researchers, developers, founders, and advocates of open-source technology; it felt more like a reckoning than a typical launch-day complaint.
Our Anthropic overlords deciding which prompts the peasants are allowed to use. pic.twitter.com/08YCSJcYSc
— Bojan Tunguz (@tunguz) June 10, 2026
High Token Consumption
The initial concern users raised was not related to safety features. Fable 5 charges $10 for every million input tokens and $50 for output tokens—this is double the cost associated with Claude Opus 4.8.
It's enjoyable, but also very token-hungry. My wallet can hear it thinking. pic.twitter.com/xgv8N5fmUO
— edisonZ (@ai_edisonZ) June 10, 2026
This pricing is already steep, but the real issue lies in how the model operates within subscription plans. Fable 5 counts usage at double the rate compared to Opus, meaning that tasks performed with Fable diminish your plan's allowance at twice the speed before incurring any API charges.
In practice, the situation deteriorated further. In our own brief testing, Fable consumed our entire daily quota with just one prompt. Even users with substantial budgets faced challenges; Bleeping Computer also tested Fable and found it depleted a $100 Max subscription's daily limit in just under nine minutes.
Scrimba’s CEO, Per Borgen, shared publicly, "Just tried Fable. It burned 1.3M tokens in 7 minutes. That's $160 per hour. Equivalent to a $333k/year salary," he posted on X.
Theo from T3 Chat tweeted that he spent over $1,000 on tokens in one day using his $200 subscription plan. Josh Ellithorpe, CTO at Pixelated Ink, stated Fable 5 “burns tokens like no other model,” leaving him with very few prompts before exhausting his quota. “Can't even review this, since my testing is so limited,” he lamented.
According to Anthropic, the Workflow mode—the feature that consumes tokens most aggressively—splits complex prompts into parallel sub-agent tasks, which inherently requires more compute power.
Additionally, a new system prompt, which is about 120,000 tokens long, is included with every new conversation. For context, this is approximately the same token context window that GPT-4o could handle before failing.
The company claims that Fable 5's efficiency per task is better than it appears on a per-token basis since it generates more comprehensive output with fewer iterations. While this may hold true in controlled tests, users with fixed daily limits found that it functioned as a budget-consuming machine.
Now that you've run out of mythos after 1 token, you can give some open, unrestricted models a try haha https://t.co/FcCjffocY7
— Teknium 🪽 (@Teknium) June 9, 2026
Hidden Performance Degradation
the claude fable 5 nerf for AI research has induced the angriest reaction from AI researchers that I've ever seen in my life
Dear Anthropic, your broke our trust and I don’t think you’ll ever get it back.
My tokens will no longer fly your way.
Mikel Artetxe, co-founder of Reka AI, also criticized this decision:
"Brilliant idea! Next up: Apple randomly reboots your Mac if you're building competing tech, Gmail silently edits your email if you mention rival platforms, and Tesla Autopilot swerves if it detects you're working on self-driving cars. All in the name of safety, of course," he tweeted.
The researchers most affected were not large labs with proprietary infrastructure but rather academics, startups, and independent developers who relied on Claude as a public tool—precisely those whom Anthropic's safety rhetoric claimed to protect. AlphaXiv, an open research platform, deemed the practice a precedent that "is not safety," asserting that safety policies should be transparent and auditable.
As believers of open research, we are disappointed to see Anthropic silently degrading Fable 5 for AI development
"Any topic related to building pretraining pipelines, distributed training infrastructure, or ML accelerator design... may have limited effectiveness through Claude… pic.twitter.com/ELE8lqQWaF
— alphaXiv (@askalphaxiv) June 10, 2026
Nathan Lambert, who recently joined Arcee AI after working with the Allen Institute, put it plainly: “To me this paints Anthropic clearly as anti science, and therefore anti progress and anti safety,” he wrote.
I don't really want to have to go to bat against Anthropic, but they've just been unnecessarily antagonistic to all of China, then not so subtly to open weight models, and now more broadly open AI research. What's next on the list?
— Nathan Lambert (@natolambert) June 9, 2026
A pseudonymous user, "CalleBTC," an AI and crypto developer who had been anticipating Fable to assist in training a world model, expressed similar frustrations: "Anthropic has lost the plot. I was literally waiting for Mythos to help me train a world model. Instead, they chose to cuck their model to stifle their competition,” he stated, calling the move “deeply unethical and disrespectful to developers and scientists."
Overall, researchers contend that Fable’s limitations may extend beyond specific topics and could be influenced by how the model categorizes users.
Matt, I can’t even say “hello” to Fable 5 except in incognito mode (memories off), because it knows I am a biomedical researcher!
It would be nice not to ban biomedical scientists before talking access. Isn’t your comment ironic? Let’s first see if you can fix punishing us! https://t.co/mBYSu9tNIb pic.twitter.com/IKI9ksBBNL
— Derya Unutmaz, MD (@DeryaTR_) June 10, 2026
Data Retention Issues
The third concern was most relevant to enterprise users, but its implications affected all users. According to Anthropic's own announcement, all interactions with Mythos-class models—Fable 5, Mythos 5, and any future models of similar capabilities—are subject to a mandatory 30-day data retention policy across all platforms where these models are available, including third-party platforms like AWS Bedrock and Google Vertex AI.
The company claims that this data will be deleted after 30 days in "almost all cases."
The issue for enterprise users lies not in what Anthropic asserts it will do but rather in the structural requirements of the policy. Companies handling sensitive legal communications, healthcare records, or confidential source code may face significant risks if they utilize these models. Users argue that existing privacy agreements with Anthropic should be revised to ensure confidentiality.
This compliance challenge is also geographical. European companies bound by GDPR's data minimization principles, or any organization needing demonstrable zero-retention for regulated workflows, are effectively barred from using Fable 5 until Anthropic provides an exception. Pseudonym X user, Lisan al Gaib, a notable figure in the AI community, pointed out this consequence:
"Anthropic just delegated a lot of European companies to the permanent underclass. If Anthropic saves data for Claude Mythos and Fable 5 for 30 days, then all companies that require zero data retention simply can't use them."
Anthropic just delegated a lot of European companies to the permanent underclass
if Anthropic saves data for Claude Mythos and Fable 5 for 30 days, then all companies that require zero data retention simply can't use them
— Lisan al Gaib (@scaling01) June 10, 2026
Clement Delangue, CEO of Hugging Face, framed the week's events within a broader context:
"Concentration of power, capabilities and economic wealth is the biggest risk in AI,” he wrote. “We need open science and open-source more than ever!"
Another user commented: “All jokes aside, it's very clear that Anthropic is the direct path to the worst type of dystopia. Their CEO is against the very technology he creates. Restricting knowledge and education on ML related topics is beyond despicable.”
Fable 5 is available free on Pro, Max, Team, and Enterprise plans until June 22. After that, it will transition to usage credits only—API rates, without any subscription coverage—with Anthropic stating it will restore broader access "as soon as capacity expands."