Anthropic's Mythos-class model allegedly leaked on Hugging Face
A purported leak of Anthropic's unreleased model sparked debate over authenticity amid skepticism it's a malware distribution scheme.
What to know
- An 8TB model claiming to be Anthropic's Mythos-class architecture appeared on Hugging Face under an anonymous account, with the original announcement deleted for malware concerns.
- Technical community consensus leaned strongly toward the leak being fake, citing a pattern of similar hoaxes, technical implausibilities, and the impracticality of the model size for casual users.
- Debate centered on whether this was a malware distribution scheme, a troll, or legitimate leak—with commenters noting historical precedent for fake high-profile content uploads.
The dispute Whether the leak is a malware hoax or merely a prank/trolling attempt, and whether the prohibitive model size actually undermines the malware hypothesis. · positions read across 10 posts and comments
The leak is almost certainly fake, likely designed to distribute malware or troll users.
-
“This is almost certainly fake. A few days ago there was supposedly a Fable 5 checkpoint 'up for sale on a deep web forum', now there is supposedly a Mythos-class model leaked on Hugging Face.”
ripbozo · Hacker News ↗
The 8TB model size makes malware distribution impractical, reducing the plausibility of a malware hoax vector.
-
“Not saying if it's fake or not cause I have no idea, but the model is almost 8 terabytes so it's not something a casual, less knowledgable user is going to download and run on a whim.”
mjr00 · Hacker News ↗
Anthropic AI company whose model was allegedly leaked
How it unfolded 4 developments, newest first · click a bar or a number to jump articlespostscomments
-
4
User points to Hugging Face discussion confirming fakeness
A commenter linked to a Hugging Face discussion thread indicating the leak was confirmed fake, while another noted the safetensors format download posed no malware risk.
“Back in the day people used to upload new movies on public torrent sites that turned out to be something completely different in order to troll people. This is probably the same.”
— cbg0, HN commenter · source -
Back in the day people used to upload new movies on public torrent sites that turned out to be something completely different in order to troll people. This is probably the same.If not, this is one hell of a way to save their IPO.
1 more of the top 2 · 3 posts in this stretch
-
Link https://huggingface.co/burner22/numbat, there is no malware concern if only safetensors is downloaded.
-
-
3
Commenters examine technical details for authenticity markers
Technical analysis of training logs and model specifications emerged. One commenter found the metrics and log lines "pretty legit at first" but remained skeptical overall, while another noted the 8TB size would be an impractical vector for casual malware distribution.
“Looks pretty legit to me at first. May be hard to run, I don't have 8TB of RAM.”
— _ache_ -
So the idea would be to use some existing model architecture and model that has been poisoned that will actually be parsable by existing inference software to try to do malicious things if it's run in a harness that allows tool use/etc? On the assumption that the type of person to download and try to run inference with the model would be smart…
2 more of the top 3 · 6 posts in this stretch
-
Looks pretty legit to me at first. May be hard to run, I don't have 8TB of RAM.The log line is also interesting.> [2026-10-08T13:51:36Z step 1830001] loss=1.0550 lr=7.19012e-05 lr_phase=decay grad_norm=1.304 tokens/s=12.03M mfu=35.1% moe_balance_loss=0.0217 expert_gini=0.0231Last line of log:> [2026-10-08T14:37:21Z step 1830400] loss=1.0421…
-
Not saying if it's fake or not cause I have no idea, but the model is almost 8 terabytes so it's not something a casual, less knowledgable user is going to download and run on a whim.
-
-
2
Community voices skepticism, flags pattern of fake leaks
Commenters quickly questioned the leak's authenticity. One noted a similar claim about a Fable 5 checkpoint circulating days earlier, suggesting a pattern of social engineering to distribute malware. Others compared it to historical tactics of uploading fake content on torrent sites.
“This is almost certainly fake. A few days ago there was supposedly a Fable 5 checkpoint 'up for sale on a deep web forum', now there is supposedly a Mythos-class model leaked on Hugging Face.”
— ripbozo -
This is almost certainly fake. A few days ago there was supposedly a Fable 5 checkpoint "up for sale on a deep web forum", now there is supposedly a Mythos-class model leaked on Hugging Face. This is just a way to get less knowledgeable people to run code (as in malware) on their machines.
-
-
1
Alleged Mythos-class model surfaces on Hugging Face
A post claiming Anthropic's Mythos-class model had leaked appeared on Hacker News, with the original URL subsequently deleted for malware concerns. The model was hosted under a "burner22" account on Hugging Face.
-
1 outlet first by HN Frontpage, 10d ago · read ↗
-
What people are saying 3 voices from 1 site · best of 10 · verbatim
- Is this the same pattern as the claimed Fable 5 checkpoint leak from days earlier?
- What legal recourse does Anthropic have against model weight leaks given that data cannot be copyrighted?
- Nov 1
-
No doubt it's fake, but 8 TB of weights would be a pretty crazy way to package malware
-
Not a lawyer, but I don't see what protection they have other than trade secrets since the weights are just data and data can't be copyrighted.
-
Naturally. There was a fake Claude 2.1 leak a couple years ago, which turned out to be a renamed open model. Maybe also others.