Modern Creator
Nate Herk | AI Automation · YouTube

Is Claude Mythos Coming?

A 12-minute reality check on the Mythos API leak and why one buried Anthropic sentence beats all the hype.

Posted
yesterday
Duration
Format
Talking Head
educational
Views
24.5K
593 likes
Big Idea

The argument in one line.

Anthropic stated in April that they do not plan to make Mythos preview generally available, nothing has reversed that, and a brief API leak is marketing noise not a launch signal; GPT-5.6 is the only real wildcard that changes the math.

Who This Is For

Read if. Skip if.

READ IF YOU ARE…
  • You follow AI model releases closely and wanted a grounded take on the Mythos API leak without the hype.
  • You use Claude professionally and want to understand what the next frontier model might look like or cost.
  • You are curious about how Anthropic and OpenAI tactically time releases against each other.
  • You want to understand what Project Glasswing is and why Mythos is gated to cybersecurity partners.
SKIP IF…
  • You are looking for confirmed insider information about Mythos release dates.
  • You are not interested in AI industry dynamics and just want a tutorial or how-to.
TL;DR

The full version, fast.

Claude Mythos is a new Anthropic model family above Opus 4.8, gated to vetted cybersecurity partners via Project Glasswing since April. A Mythos identifier appeared briefly on the public API then vanished, triggering launch speculation. Anthropic stated in April they do not plan to make the preview generally available, and nothing since has changed that. The most probable outcome is that Mythos capabilities quietly fold into a future Opus release. The one force that could accelerate a public launch is competitive pressure from OpenAI dropping GPT-5.6, which betting markets put at 80-plus percent odds for June 2026.

Free for members

Chat with this breakdown — free.

Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.

Create a free account →
Chapters

Where the time goes.

00:0000:27

01 · Intro and hook

Mythos API identifier appeared and vanished this morning. Promise to cut through the hype with one buried Anthropic quote.

00:2701:51

02 · What Mythos Actually Is

Model family above Opus 4.8. Cybersecurity is the headline capability. Gated to vetted partners via Project Glasswing. Public benchmark was close with an existing GPT model.

01:5102:59

03 · The Case It Is Coming Soon

Glasswing expanded from 50 to 150-plus orgs. Betting markets at two-thirds odds by July. Access circle keeps growing.

02:5904:51

04 · Why He Is Not Buying It

Anthropic own April statement still stands: no plans for general availability. The morning leak is likely intentional marketing. A rough leak travels farther than a launch video at zero cost.

04:5106:52

05 · The Forces Behind the Hype

Force 1: IPO timing, S-1 filed at 965B, scary-model narrative is useful before selling stock. Force 2: Irony of calling for a global AI pause while expanding Mythos access.

06:5208:12

06 · The OpenAI Factor

GPT-5.6 at 80-plus percent odds for June. Anthropic and OpenAI documented pattern of timing releases against each other. Dropping Mythos same day GPT-5.6 lands would flood the feed.

08:1210:11

07 · Where He Lands

No public Mythos button anytime soon. Capabilities fold quietly into the next Opus. OpenAI move is the one variable that changes the math, but even then he bets on a new Opus not public Mythos.

10:1112:13

08 · 3 Cases: Bull, Base, Bear

Bull: limited gated paid Mythos this year at 5x Opus pricing. Base: capability folds into Opus, full Mythos much later. Bear: Mythos stays locked to trusted partners effectively forever.

12:1312:24

09 · Final Thoughts

Watch the next Opus, not the Mythos name. GPT-5.6 is the real launch window signal.

Atomic Insights

Lines worth screenshotting.

  • Anthropic stated in April they do not plan to make Mythos preview generally available and nothing since has walked that back.
  • A rough API leak travels farther than a polished launch video and costs the company nothing to execute.
  • Project Glasswing expanding from 50 to 150-plus organizations is access widening inside a gated program, not a path to public availability.
  • Betting markets put GPT-5.6 at 80-plus percent odds for June 2026, higher confidence than Mythos going public by July.
  • Anthropic and OpenAI have a documented pattern of timing releases to land on top of each other within minutes or hours.
  • The strongest reason to drop public Mythos now is to overshadow GPT-5.6 on launch day, not safety readiness.
  • Glasswing partners currently pay $25 per million input tokens and $125 per million output tokens, roughly five times Opus pricing.
  • The one public Mythos benchmark showed it neck-and-neck with a GPT model already publicly available, undercutting the scariest-model-ever framing.
  • The base case is not a Mythos launch but Mythos capabilities folding quietly into the next Opus while most users never realize it.
  • IPO timing creates a structural incentive to keep the scary-model narrative alive regardless of actual release plans.
  • A company can call for a global AI pause treaty while expanding access to its most capable model; the two positions are uncomfortable together but not logically contradictory.
  • When two AI labs are racing, watch the competitor launch window as a more reliable signal than leak screenshots.
Takeaway

How to read an AI model leak without getting played.

WHAT TO LEARN

A leaked API identifier is not a launch, and three structural forces explain why the hype machine runs anyway.

  • Check the primary source first: Anthropic own April statement said they do not plan to make Mythos preview generally available, and nothing since has reversed it.
  • A rough leak spreads farther than a polished launch announcement at zero cost, which means whether intentional or not, the outcome is free earned media.
  • Access widening inside a gated program is not the same as public access; 150 vetted partners is still a controlled rollout.
  • Prediction markets are real signal but separate the two bets: GPT-5.6 at 80-plus percent for June is not the same as Mythos going public by July.
  • IPO timing creates a structural incentive to keep a scary-model narrative alive regardless of actual release plans.
  • The base case for any frontier model launch is that capabilities fold quietly into the next incremental release and most users never see a branded new model.
  • Competitive timing between AI labs is documented and real: watch the competitor launch window as a more reliable signal than leak screenshots.
Glossary

Terms worth knowing.

Claude Mythos
A new Anthropic model family positioned above Claude Opus 4.8, described internally as their most powerful model ever. Currently gated to vetted cybersecurity partners and not publicly accessible.
Project Glasswing
Anthropic invite-only program that deploys Mythos to vetted cybersecurity organizations, tasking the model with finding and patching security vulnerabilities in critical software.
GPT-5.6
OpenAI anticipated next frontier model, with prediction markets assigning 80-plus percent odds of a June 2026 release, widely seen as the competitive trigger that could force Anthropic hand on Mythos timing.
S-1
Registration statement a company files with the SEC before going public. Anthropic confidentially filed one in early June 2026 at a reported 965 billion dollar valuation.
Bull/base/bear case
Framework borrowed from financial analysis mapping the optimistic, most probable, and pessimistic scenarios for a given outcome, used here to map three possible Mythos release trajectories.
Resources

Things they pointed at.

01:51productProject Glasswing
04:57linkAnthropic S-1 confidential filing June 2026
05:13linkAnthropic AI safety report When AI builds itself
00:00productUppitai agency
Quotables

Lines you could clip.

00:13
There is one line buried in Anthropic own writing that flips this whole story on its head.
Tight curiosity-gap setup that works standalone as a short-form hook.TikTok hook↗ Tweet quote
03:20
We do not plan to make Mythos preview generally available. That was literally their own words, not mine.
Direct primary source quote that is the thesis of the whole video.IG reel cold open↗ Tweet quote
08:38
A leak does not change the plan.
Six words, standalone, works as a text card.newsletter pull-quote↗ Tweet quote
05:48
Hype has a motive right now, and that is definitely worth keeping in mind.
Evergreen framing for any AI news cycle.TikTok hook↗ Tweet quote
The Script

Word for word.

Read-along

Don't just watch it. Burn it in.

See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.

metaphor
00:00Okay. So this morning, Mythos showed up on Anthropic's API for a few minutes, and then it vanished. So, naturally, people screenshotted it, and my whole feed just lost it.
00:09You know? It's dropping. Next week, it's over.
00:12So before we start counting down to launch day, there's one line buried in Anthropic's own writing that flips this whole story on its head. So what I wanna do today is catch you guys up on what Mythos actually is, what really happened this morning, and my honest bet on whether we'll ever get our hands on it. So let's get into it.
00:27So real quick, in case you haven't been following this whole Mythos saga, here's pretty much the short version. Mythos is basically the tier above Opus. So right now, the best model that you can actually log in and use is Cloud Opus 4.8.
00:38So Mythos apparently sits right on top of that. And in their own leaked draft, they called it a step change and the most powerful model they've ever built. So it's not like we're jumping from OPUS 4.7 to OPUS 4.8.
00:49This is an actual new model family entirely. The thing that it's really, really weirdly good at is cybersecurity.
00:56So being able to find security holes and patch them, which is exactly why it's also a little scary because the same skill that finds holes so that you can fix them is also the same skill that can find holes so that you can break in and hack things. And that's exactly why the only people who have actually gotten to use Mythos so far are just a handful of big cybersecurity companies.
01:16Anthropic basically started rolling it out to them through a program called project Glasswing to point Mythos at critical software and have it hunt down these security holes and patch them, and reportedly, it's really, really good at that. It's been able to dig up some pretty serious bugs in super old widely used code that had just gone unnoticed for years.
01:34Though, to be fair, on the one public benchmark that we actually got, it was kind of neck and neck with a GPT model that we can already use publicly. So them saying, you know, scariest model ever might be a little oversold. It might be a little bit of, you know, marketing and IPO hype, but we don't truly know.
01:49So, anyways, that's what Mythos is. The real question is what actually happened this morning and why it sends everyone into a frenzy. So today, a Mythos identifier showed up live on the API.
01:59People grabbed their screenshots, and then the identifier got pulled. And the case that it's coming soon, it's actually not crazy because if you look at how things have been rolling out, Mythos started out going to around 50 partners back in April when project Glasswing was announced.
02:13And then just a few days ago, that number jumped from 50 to a 150 organizations across more than 15 countries. So the circle of people getting access keeps getting bigger. You know, it's growing pretty fast.
02:25And then there's also things like the betting markets. You know, people are literally putting money down on when Mythos will go public. And the number that keeps getting passed around is somewhere around two thirds odds that it shows up by the end of July, so, you know, within the next two months, and those odds have been climbing.
02:40So this isn't just a couple of hype accounts on x that are yelling and saying, hey. You know, blah blah blah. Here's what I think.
02:48There's actual money that's also riding on it, which is pretty interesting. But if you look at only the leak plus the Axis widening plus the betting markets, you'd think, yeah, it's definitely coming soon. And that's the honest version of the hype because there is kind of a real case.
03:03But let me talk about why I'm not buying into it because it all comes back to what Anthropic actually said. So Mythos was never, you know, a public product.
03:11It went out through Glasswing, which was invite only, so just the vetted partners and the government, and they locked it down on purpose because a model this good at hacking is a weapon if it's in the wrong hands. Right? So back in early April when Anthropic straight up said, we do not plan to make mythos preview generally available.
03:28That was literally their own words, not mine, and it's the preview that we're talking about here. So they're not promising this kind of power never reaches you. They're just saying that this one gated version, the thing that everyone's losing their minds over right now isn't what's going public, at least not yet.
03:43Now, yeah, once again, to be fair, that statement is back from, you know, April, so a couple months ago, but nothing since has walked it back. And the leak this morning isn't Anthropic opening the doors either. But one thing that I have to be honest about that I am pretty skeptical about is that this leak is intentional or past leaks have been intentional because it's honestly a genius marketing play.
04:03You know? You do a quick leak.
04:05You don't actually have to release anything valuable. You pull it back in, and then you just, you know, start eating your popcorn and watching the show Because now you have all of these accounts on x that are jumping all over it. Everyone's talking about it.
04:16It's trending on Twitter. It's gonna trend on LinkedIn, and you've got all these people talking about your company, and you didn't have to do anything.
04:23You hardly had to put any time. You had didn't have to spend any money. It's pretty genius if you think about it.
04:29So I do think sometimes, not only Anthropic, I think that in general, that's kinda just the culture now is sometimes people do things like that to build hype and anticipation because it sticks out different than if they were to put their own, like, hype video out. You know? So if Anthropic flat out said they're not releasing it, then why does half the Internet still think it's coming?
04:47Because there are some other forces that are pushing on this that we definitely want to address. So I think the first one is money. You know?
04:55And the timing here is pretty incredible. On June 1, so at the time of recording this video about six days ago, Anthropic confidentially filed to go public at a valuation of around $965,000,000,000.
05:08So basically right there knocking on the door of a trillion dollars. So they're about to sell stock to the public, and then three days later, they drop this big dramatic report basically saying AI is starting to improve itself. I just did a video about this yesterday, and they basically said the world might need to think about hitting the brakes and pausing.
05:24So if you put those two things right next to each other, you're about to raise a huge giant pile of money. And right before that, you remind everyone that you're sitting on the most powerful, really scary AI model in the world, and that's a really good story to be telling while you're about to sell stock.
05:40I'm not saying the tech isn't real. I do think it is. But hype has a motive right now, and that's definitely worth keeping in mind.
05:46So that's force number one. Force number two is that the report itself, you know, the irony, it's too good to skip over.
05:53Because in that report, Anthropic is literally calling for the option to slow down frontier AI, like a global agreement, like a treaty, where everybody can verify that everyone else is slowing down as well, which, you know, it's a reasonable thing to ask for. I do think that there's the speed is kind of scary and there's a huge gap and it keeps getting wider and wider.
06:13You know, you're telling the world, hey. Maybe we should all consider pausing while you're quietly sitting on the scariest model anyone's ever built, and you're handing it out to more and more people every single week and teasing that it's gonna go public. You know, even some of the big leak accounts kind of looked at that and went, yeah.
06:26That's a little bit contradictory. And, I mean, I completely agree. But to be fair, once again, Anthropic stance has actually been pretty consistent in basically saying, hey.
06:34We'll slow down if everyone slows down together, but we're not just going to unilaterally, like, give up our lead right now and hand that to someone else, which also is fair. So it's not just it's not just pure hypocrisy, but you can believe all that, and you can still admit the timing is a little bit weird.
06:50Right? Okay. So force number three, though, and this is probably the biggest one, is OpenAI.
06:56Because the other reason everyone expects Synthropic to make a move is that OpenAI's next model, the one that everyone's predicting is gonna be GPT 5.6, is supposedly right around the corner. And these two companies, you guys probably know this, they're you know, they tend to do things right on top of each other.
07:10They try to overshadow and stack on top. And that's just that's not me guessing. You know, we've seen these patterns.
07:15Like, earlier this year, OpenAI shipped a new coating model within minutes of Anthropic shipping one. And, you know, on their pricing just last month, OpenAI rolled out, like, the two free months of Codex to get people to switch over, and then Anthropic turned around and bumped everyone's clawed limits by 50% that same day or something like that.
07:30The point I'm trying to make is that they watch each other closely, and neither one wants to get overshadowed, and they want to overshadow the other company. And those same betting markets that we talked about earlier, they actually like g b t five point six's odds even more than Mythos.
07:44So people are betting that it lands by the end of this month in June, and the odds on that are up, I think, north of 80 now. So people are even more sure that OpenAI is going to ship a new model first. So the bet that a lot of people are making is basically g p t 5.6 drops, and then Anthropic fires back within days so they don't get buried, maybe even same day.
08:03The only real question is what does Anthropic actually fire back with? Is that maybe an opus five or 4.9, or is that like the big daddy mythos?
08:12So is Claude Mythos coming? Alright.
08:15Here's where I land on it. And, obviously, I could be completely wrong. I don't have any insider information.
08:19A public model that you can just log in and use called Mythos, I don't think that's happening anytime soon. And it might not ever happen under that name, at least not the way that we're picturing it right now.
08:30I mean, they kind of told us the plan. A leak doesn't change the plan. So, I mean, everyone waiting for a big Mythos button to show up in the app next week, I think they're going to be waiting a while.
08:40So what do I actually think happens? I think the capabilities start to show up, but kind of quieter. You know, I think the stuff that makes Mythos Mythos, the stuff that makes it so special, those features sort of start to get baked into future versions of Opus.
08:54And once Anthropic decides the safety guardrails are solid enough, then that's when they we start to see some more stuff. Right?
09:01I'm not sure if we'll get the model called Mythos or, you know, or maybe when we do get Mythos, it's not gonna be the full most powerful one. It'll be kind of toned down. Now the one thing that could really change that math is the OpenAI race because the single best reason for Anthropic to drop Mythos soon would be timing it to land right on top of GPT 5.6.
09:19If OpenAI shipped that, and then same day or a day later, Anthropic drops Mythos, the whole feed is suddenly talking about Mythos. YouTube, LinkedIn, x, whatever you're on, everyone's testing Mythos, and GPT 5.6 basically launches into an empty room. Now people obviously would still come around, and if they're already loyal to codex, they would use 5.6.
09:38So it wouldn't get destroyed, but all of the hype on that one or two days or maybe for the first week would all be going towards anthropic, which is obviously a massive win and another reason why I think that the mythos timing might be glued to whenever OpenAI finally makes their move. But even with that temptation sitting right there, my money is still on the thing they actually dropped being a new opus, not a public mythos.
09:59Now in the long run, I do think that the capability itself probably does reach the general public eventually, so the crowd isn't totally wrong in thinking that. I think that it's just too early, and I think that they're too hung up on the name mythos.
10:11So let me give you three ways that this could potentially play out so that you guys can kind of, like, place your own bets. But we've kinda got the bull case, right, which is very exciting. It's competitive pressure plus all that IPO momentum pushes Anthropic to put out some limited, paid, still kind of gated Mythos later this year.
10:30And even then, I'd bet it's a toned down version, not the full power of Mythos they're sitting on so that they can keep a 5.1 and a 5.2 ready to roll out as soon as GPT drops something else. And if the preview price is any clue, touching it won't be cheap.
10:43Right now, Anthropix charging Glasswing partners around $25 per million input tokens and 125 for output, which is five times what Opus cost you today, which I've seen already so many comments about how expensive Opus already is. So that's the bullish case.
10:59The base case, which is the one that I'd probably put my money on right now, is that no public mythos. The capability just folds into the next opus sometime in the back half of this year or early next, and most people never even realize that that's happening. And then we get Mythos way down the line once people are feeling a lot more secure, once more partners through Project Glasswing have gotten this model and played around with it and things have been locked down, then the public gets Mythos.
11:24And then we have the bear case, which is that Mythos just stays locked up pretty much forever or it's you know, you have to be a trusted partner in order to get access to Mythos, like, through project Glasswing, and it stays more of a defensive security tool for these vetted partners forever. Then And the public ends up just getting a watered down slice of some of the features inside of it, maybe inside Opus or something.
11:45And given everything they've said about how dangerous it is, that one's not a crazy theory either, at least I don't think. So, anyways, zooming out, I think that Mythos is real and very powerful, and it's almost certainly not the big public launch that the timeline seems to think it is right now, at least not yet. So the thing that you should actually be waiting for isn't a big Mythos launch.
12:04It's probably the next Opus model and whether OpenAI ships g b t 5.6 and if that forces Anthropic's hand to drop something else pretty quick after that. But, anyways, that's gonna do it for this one. So if you guys enjoyed the video or you learned something new, please give a like.
12:18It definitely helps me out a ton. And as always, I appreciate you guys making it to the end of the video, and I'll see you all in the next one. Thanks, guys.
The Hook

The bait, then the rug-pull.

For a few minutes on the morning of June 6, a Mythos model identifier appeared live in Anthropic API. People screenshotted it. The feed lost its mind. And then it vanished. Before anyone starts counting down to launch day, there is one sentence buried in Anthropic own writing from April that reframes the entire event and it has not been walked back.

Frameworks

Named ideas worth stealing.

10:11model

Bull / Base / Bear Case

  1. Bull: limited gated Mythos late 2026
  2. Base: capability folds into next Opus
  3. Bear: stays locked to Glasswing partners forever

Three probability-weighted scenarios for when and if Mythos reaches the public, borrowed from financial analysis framing.

Steal forAny video where you want to map a spectrum of outcomes without committing to one prediction
04:51list

Three Forces Behind the Hype

  1. Money and IPO timing
  2. Irony of the pause call
  3. OpenAI competitive pressure

Structural forces that amplify Mythos hype independent of whether a public launch is actually planned.

Steal forExplaining why a narrative exists even when the underlying facts do not support it
CTA Breakdown

How they asked for the click.

VERBAL ASK
12:13subscribe
If you guys enjoyed the video or you learned something new, please give a like.

Standard like-and-subscribe close after all content is delivered. No mid-roll CTA.

FROM THE DESCRIPTION
Storyboard

Visual structure at a glance.

open
hookopen00:00
model ladder
valuemodel ladder00:27
Project Glasswing
valueProject Glasswing01:09
access expanding
valueaccess expanding02:05
manufactured leak mechanic
valuemanufactured leak mechanic04:04
GPT-5.6 odds comparison
valueGPT-5.6 odds comparison07:40
no public Mythos button
valueno public Mythos button08:41
pricing card 25 in 125 out
valuepricing card 25 in 125 out10:37
close and CTA
ctaclose and CTA12:13
Frame Gallery

Visual moments.

Watch next

More from this channel + related breakdowns.

Chat about this