A new Claude model just finished training.What's real, what's rumor, and what it means for you.

First, the model that started all this. Fable 5.

✓ CONFIRMED

On June 9, Anthropic released two models at the very top of its line-up: Claude Fable 5 and Claude Mythos 5.

They sit in a tier Anthropic calls Mythos-class — a step above the Opus class. In Anthropic's words, Mythos-class models "sit above our Opus class in capability." Fable 5 is described as state-of-the-art on nearly every benchmark, strong at software engineering, knowledge work, vision, and science.

Here's the part most people miss — Fable 5 and Mythos 5 are the same underlying model. The difference is the safety settings.

So Fable 5 is the one the public could use — with guardrails. When you ask it about cybersecurity, biology, or chemistry, it quietly hands off to Claude Opus 4.8 instead. Mythos 5 is the same model with those guardrails off, locked to vetted partners through a programme called Project Glasswing — cyber-defenders and select biomedical researchers.

The confirmed specs on Fable 5, for the record:

June 12: the US ordered it switched off.

✓ CONFIRMED

Three days after launch — at 5:21pm ET on June 12 — Anthropic received an export-control directive from the US government.

The order: suspend all access to Fable 5 and Mythos 5 for any foreign national, anywhere in the world, inside or outside the US — including Anthropic's own foreign-born staff.

There's no way to half-comply with that. So Anthropic disabled both models for everyone. Every other Claude model stayed up.

Anthropic said it plainly ↓

The stated reason was national security — the government believed someone had found a way to "jailbreak" Fable 5. Anthropic's own review found the technique was, in effect, asking the model to read a codebase and fix its flaws — something Anthropic says is widely available from other models already.

Anthropic pushed back publicly. Their position, in short: perfect jailbreak-resistance isn't possible for any model provider, and if that were the bar, no new frontier model could ever ship.

That last line matters for you, and we'll come back to it.

And then — a new model appeared. Quietly.

~ RUMOR — X posts only

This is the part everyone's talking about. It is also the part that is not confirmed by Anthropic. So rather than paraphrase, here are the actual posts — read them yourself and judge.

The core claim — a new, more capable Mythos has finished training ↓

The concrete signal — a "claude-sonnet-5" slug on an Anthropic partner provider ↓

More from the AI-news feeds ↓

And the theory going around — that the embargo freed up the compute to train it ↓

Notice the two separate signals here. One is a new Mythos-class model (the top tier). The other is a Sonnet-tier slug (the everyday workhorse tier). They might be the same launch wave. They might be two different things. Nobody outside Anthropic knows yet.

And one important check I did for you — I looked at Anthropic's official model docs. As of right now, there is no claude-sonnet-5, no Mythos 5.1, and no Mythos 6 listed. So treat all of this as smoke, not fire. Real smoke, from credible watchers — but smoke.

So is it Mythos 5.1, or Sonnet 5? Let's reason it out.

Here's the thing almost every thread is getting wrong: there are probably two different models in this story, not one.

Andrew Curran says a new Mythos-class model finished training — that's the top tier, above Opus. But the slug that was actually spotted is claude-sonnet-5 — that's the Sonnet tier, the everyday workhorse, a rung lower. A single model can't be both. So the smart move is to hold two possibilities side by side.

Quick map first, then the reasoning under each:

One last thread worth naming: the speed. Some watchers think the export order, by cutting off a huge chunk of usage, freed up the compute Anthropic poured into training this one. That's a theory, not a fact — but it's a tidy one, and there's real competitive pressure behind it, with China's GLM-5.2 pushing hard at the frontier.

Own the system. Not the model.

Step back from the rumor for a second, because the real lesson here has nothing to do with a model name.

On June 9, Fable 5 was the most powerful model on earth. Three days later it was gone for most of the planet — not because it got worse, but because of a force completely outside anyone's control.

So look at who got hurt, and who didn't.

The people who built their work directly on Fable 5 — hardwired to that one model — woke up stranded. Their tools just stopped. Anyone outside the US, especially.

The people who built on a system — a model-agnostic Agent OS holding their memory, their workflows, their agents — didn't break stride. The moment Fable 5 went dark, they pointed the same system at the next best model and kept shipping. Same memory. Same prompts. Same agents. Different engine underneath.

And here's what flips it from "survived" to "won": a system that always points at the best model available quietly gets better. When the new model lands, they upgrade by changing one setting — while everyone else rebuilds from scratch.

That's the whole shift. The model is a rented engine. It changes every few weeks, and it can be pulled, priced up, or switched off by people who aren't you. The thing you actually own — the durable asset — is the system around it.

Three practical takeaways. No hype.

The whole story in six lines.