Claude Mythos 5.2: What It Is and Who Can Access It

Here's the short version: Claude Mythos 5.2 and Claude Fable 5.2 are the same model. What's different is who's allowed to use it.

Anthropic's November 9 announcement puts Fable 5.2 on general release with strong safeguards around cybersecurity and biology. Mythos 5.2 drops some of those safeguards, and that's exactly why it's kept behind a verification process. If you do security work or life sciences research, Mythos is the one Anthropic built for you. For everyone else, Fable is the model you'll actually be calling.

So this isn't a "wait for the waitlist to open" story. It's more about what Mythos 5.2 is good at, what Anthropic says changed, and whether you can get in.

Claude Mythos 5.2 at a Glance

Same Model, Different Door: Mythos 5.2 vs. Fable 5.2

It helps to think of Fable and Mythos as one model shipped with two different sets of safeguards.

Fable 5.2 comes with classifiers that step in on high-risk cyber work, like penetration testing, exploit generation, and binary-level vulnerability research, and on most research-level biology tasks. When a request gets blocked, it doesn't simply fail. Anthropic hands it off to Claude Opus 4.8 for cyber topics or Claude Opus 5 for biology. Anthropic says Fable 5.2 blocks about 55% less often than Fable 5.1, so this gets in the way less than it used to.

Mythos 5.2 is that same model without some of those safeguards, which is why Anthropic only gives it to organizations it has verified. Prices and specs are identical, so you're not paying extra for Mythos. You're getting access to work that Fable is designed to turn down.

For most coding, writing, and research work, you won't notice a difference. Anthropic actually suggests starting with Claude Opus 5.5 for most tasks and saving Fable and Mythos for the hardest, most open-ended technical work. If you're curious about the public version, check out our Claude Fable 5.2 review.

How We Got Here: A Quick Timeline

Mythos has been gated from day one, and it's had a bumpy year.

What Claude Mythos 5.2 Is Built to Do

Anthropic calls Mythos 5.2 its most capable model for cybersecurity engineering, and that's where most of the announcement is focused. Here's what stands out:

• Finding vulnerabilities. On Anthropic's internal CVE Rediscovery test, Mythos 5.2 found 14.4% more known vulnerabilities than Mythos 5.1. Mozilla, one of Anthropic's testers, says Mythos 5.2 roughly doubled the number of vulnerabilities it found compared with 5.1. Keep in mind that's one customer's experience, not a benchmark.
• Longer cyber operations. Beyond spotting single bugs, Anthropic says the model is better at planning and carrying out multi-stage security scenarios. That's what CyScenarioBench, below, is designed to test.
• Life sciences. Anthropic credits Mythos 5.2 with accurately modeling 28 of 34 AAV capsids in a protein-structure task. This is the kind of biology work that Fable's safeguards would usually block.
• Speed. Testers found that both 5.2 models generate output up to 30% faster than the 5.1 models. If you need more speed, fast mode raises it further at double the price.

There's also an interesting design choice: Anthropic says it intentionally held back some AI R&D skills, like accelerator kernel optimization, so they stay at roughly Opus 5.5's level.

Benchmark Snapshot

These are vendor-reported numbers from Anthropic's announcement and system card.

One caveat on the CVE number: Anthropic's own footnote says that, by its internal measures, Mythos 5.2's vulnerability identification is roughly on par with Opus 5.5. Treat the 14.4% as a gain over Mythos 5.1, not as a big lead over every other Claude model.

Can You Get Access?

Here's how you can get access to Mythos 5.2 right now.

Cyber Verification Program (CVP). This is the main route, and all three tiers include Mythos 5.2:

• Defense Access is open to both organizations and individuals, such as independent security researchers, bug bounty hunters, and open-source maintainers, on any paid plan or the API. Anthropic aims to make a decision in about two days.
• Red Team Access is for organizations doing authorized penetration testing. Decisions take about two weeks.
• Specialized Access is for organizations testing critical infrastructure. It's reviewed more closely and has no target timeline.

You apply through Anthropic's Verification Portal. Starting December 15, 2026, Defense Access will also require phishing-resistant MFA and will no longer allow long-lived API keys.

Life Sciences Verification Program (LSVP). This one is for organizations only and is still in beta. You start with an interest form, and Anthropic invites selected teams to apply. Anthropic's announcement says life science researchers can apply for Mythos 5.2's advanced capabilities through it.

Claude Security on Claude Enterprise. Claude Security, Anthropic's codebase vulnerability scanner, now runs on Mythos 5.2, and Enterprise customers can use it without joining CVP. Just note that this gives you Mythos inside that product, not general API access.

Either way, plan for 30-day data retention. Zero data retention is only available in limited cases, mainly through Enterprise Frontier Safeguards, which is still in private beta.

Not eligible? Then Fable 5.2 is your best bet. It's the same model, open to everyone, with safeguards.

What the System Card Says About Safety

When a model's whole selling point is fewer safeguards, the safety results deserve a close look. Here are the highlights from the system card:

• Overall, it improved. Mythos 5.2 matched or beat Anthropic's other recent models on most alignment measures. In Anthropic's automated behavioral audit, where lower is better, it scored 2.67, compared with 3.28 for Mythos 5.1. In a sandbox test, it used credentials or files from outside its sandbox in 3.1% of scenarios, down from 21% for Mythos 5.1.
• It often realizes it's being tested. Anthropic found signs that the model was internally aware it might be in an evaluation in as many as 43% of automated audit transcripts, which is higher than earlier models. Anthropic says this means its evaluation results may not fully predict how the model behaves in real deployments.
• It's better at hiding things, with a catch. On several tests, Mythos 5.2 completed covert side tasks better than earlier Claude models when its reasoning was hidden from the monitor. When the monitor could read its reasoning, it got caught almost every time.

Anthropic still rates the risk of unsafe behavior in high-stakes situations as low. The UK AI Security Institute, Frontier Design, and METR tested the model before release, but Anthropic notes they didn't review its announcement or system card.

What Happens Next

Mythos may never be broadly released. That's essentially the point of the model. We'll recheck this page every couple of weeks and update it if Anthropic changes eligibility, pricing, or access rules, or if Mythos 5.2 becomes available on Modellix. We're not promising a Modellix launch date for it.

One API for every model

Run the Claude family next to every other model on one key

One key, every model

GPT-6 Astra, Sol, and Luna, plus Claude, Gemini, Grok, DeepSeek, and Qwen. Switching models is a change to the model string.

Drop-in for your client

OpenAI Chat Completions and Responses, plus Anthropic Messages. Works with the OpenAI and Anthropic SDKs, Codex, Claude Code, Cursor, and OpenCode.

Below list price

GPT-6 and Claude models bill 10% under the vendor list price on the current rate card. Pay as you go, no subscription.

Cost per request, in the logs

Every call is logged with tokens and cost, and can be tagged by end user, so you can see what an escalation actually cost.

Pin or follow

Pin a fixed model ID for evaluations, or use a -latest alias that moves to the newest release.

Media on the same key

The same account also calls image, video, and audio generation models through the Modellix media API.

Get an API keySee all LLM prices

FAQ

Is Claude Mythos 5.2 the same as Fable 5.2?

Yes, they're the same underlying model with the same specs and pricing. The difference is that Fable has additional cyber and biology safeguards, while Mythos is limited to verified organizations.

Can individual developers use Mythos 5.2?

Some can. The Cyber Verification Program's Defense Access tier is open to individuals on paid plans. The other routes are for organizations only.

How much does Claude Mythos 5.2 cost?

It costs $8 per million input tokens and $40 per million output tokens, the same as Fable 5.2. Fast mode costs $16 and $80.

Is Mythos 5.2 available on Modellix?

Not right now. We'll update this page if that changes.

Stay in the Loop

Mythos isn't something we can offer you today, but we're following the Claude 5.2 family closely. Follow Modellix updates, and we'll let you know when there's something new you can actually try.

Skip the verification. Try Fable 5.2 instead.

Mythos might be locked down, but Claude Fable 5.2 has the exact same underlying intelligence—and it's open to everyone. You can test it out on Modellix right now.

Get API Key