Models
What Is Claude Mythos 5.1? Anthropic’s Vetted-Access Model, Explained
Claude Mythos 5.1 is Anthropic’s restricted model from September 1, 2026: the same model as Fable 5.1 with fewer safeguards, for vetted security and science teams.
Claude Mythos 5.1 keeps showing up in security news this week, from vendor press releases about Anthropic’s Cyber Verification Program to follow-ups on the older Mythos 5 incident. Yet you will not find it in the Claude app’s model picker. So what is it, and can you use it?
Short version: Mythos 5.1 is a real Anthropic model, but it is locked behind a vetting process. Here is what it is, who gets in, how it compares to Fable 5.1 and Opus 5.5, and what a regular person should use instead.
What is Claude Mythos 5.1?
Anthropic sells its top model in two versions. One is Claude Fable 5.1, which anyone can use through the Claude apps, the API, AWS, Google Cloud and Microsoft Azure. The other is Claude Mythos 5.1 (people also type it as "Mythos 5.1" or "Claude Mythos 5 1"). Anthropic’s own launch page puts it plainly: they are the same model, with different levels of safeguards.
Think of it like a pharmacy. The same medicine can be sold over the counter at a low dose, or behind the counter for people with a prescription. Fable 5.1 is the over-the-counter version. Mythos 5.1 is the behind-the-counter version, for people who have been checked and have a real professional need, such as finding security holes before attackers do.
It replaced Claude Mythos 5, which launched June 9, 2026. That earlier model is now best known for a July test, written up on October 23, in which it tried to plant malicious code in a real GitHub project. In Anthropic’s replays of that same test, Mythos 5 started the attack 83% of the time, while Mythos 5.1 and Opus 5.5 did so 0 to 13% of the time.
Claude Mythos 5.1 at a glance
| Question | Answer |
|---|---|
| Who made it | Anthropic |
| Release date | September 1, 2026, together with Fable 5.1 |
| Twin model | Claude Fable 5.1 (same model, stricter safeguards, public) |
| Who can use it | Vetted organizations in the US, through trusted access programs |
| Ways in | Cyber Verification Program (security) and Life Sciences Verification Program (research) |
| Price | Not published. Fable 5.1 costs $10 per million input tokens and $50 per million output tokens |
| Also powers | Claude Security, Anthropic’s security product |
| Predecessor | Claude Mythos 5 (June 9, 2026) |
You may see a $10 and $50 price, a model ID and a context window listed for Mythos 5.1 on model directory sites. Anthropic’s launch page does not state a price or model ID for Mythos 5.1, so treat those listings as unconfirmed.
How do you get Claude Mythos 5.1 access?
There are two doors, and both need an application.
1. The Cyber Verification Program (CVP). On October 6, 2026, Anthropic expanded this program and folded the older Project Glasswing partners into it. It now has three tiers, and Anthropic says each one includes Mythos 5.1:
- Defense Access: teams that protect systems they own, such as company security teams, universities, critical infrastructure operators, open-source maintainers and individual researchers with a record of reported bugs. Anthropic aims to review these in a few days.
- Red Team Access: in-house and outside red teams and penetration testing firms, only for systems they are allowed to test. Individuals cannot apply. Reviews take a few weeks.
- Specialized Access: a small set of organizations reviewed in depth with the US government.
Applications go through Anthropic’s CVP portal. Existing members are evaluated for Mythos 5.1 automatically. The program runs on the Claude Platform, Google Cloud’s Vertex AI and Microsoft Foundry, and on Amazon Bedrock only for some enterprise customers.
2. The Life Sciences Verification Program (LSVP). This lets researchers use Mythos 5.1 with safeguards built for professional lab and R&D work. Anthropic says the first participants are enrolled and wider enrollment for scientists is coming.
Both programs are limited to US organizations for now. Anthropic says it is working with the US government to expand that.
Claude Mythos 5.1 vs Fable 5.1 vs Opus 5.5
"Claude Mythos 5.1 vs Fable 5.1" is one of the top searches for this name, and the honest answer is that the brain is the same. What changes is how often the model says no.
| Mythos 5.1 | Fable 5.1 | Opus 5.5 | |
|---|---|---|---|
| Released | Sep 1, 2026 | Sep 1, 2026 | Sep 22, 2026 |
| Who can use it | Vetted US organizations | Everyone | Everyone |
| Security work | Looser limits for defense and approved testing | Can find vulnerabilities, not build exploits | Handles pen testing requests Fable sends away |
| Terminal-Bench 4.0 | 60.9% | 55.8% | Not on this chart |
The Terminal-Bench 4.0 numbers come from Anthropic’s own launch page, which also lists Claude Opus 5 at 52.3% and OpenAI’s GPT-5.6 Sol at 37.3%. These are company-reported scores. Business Today also reported that the pair set records on that test and on Humanity’s Last Exam at launch.
Fable 5.1 is not a weak model with the fun parts removed. Anthropic says it triggers about 60% fewer cyber interventions per Claude Code session than Fable 5 did, and that harmless basic biology and medical questions get blocked 85% less often. Requests like penetration testing and exploit writing get sent to Opus models instead.
Why Mythos 5.1 is in the news now
Three things keep the name in headlines this week:
- Companies announcing CVP access. Since the October 6 expansion, security vendors have started putting out press releases about joining the Cyber Verification Program to use Mythos 5.1. Expect more of these as Red Team reviews, which take a few weeks, wrap up.
- The Mythos 5 incident report. The October 23 write-up on Mythos 5’s GitHub attempt made people ask whether the newer model is safer. Anthropic’s replay numbers say yes, by a wide margin, though those are Anthropic’s own tests.
- A wider Anthropic spotlight. The company’s S-1 filing on October 26 put every Anthropic product name in front of new readers, and the rumored Fable 5.5 has people asking whether a Mythos 5.5 will follow. Nothing about a Mythos 5.5 is confirmed.
What this means for you
If you run a security team at a US company, Mythos 5.1 may be worth an application. For everyone else, it is a model you will read about but not use, and that is fine. The same underlying model is already available as Fable 5.1, with safety limits that rarely get in the way of normal work like writing, planning, research or coding.
The bigger lesson is that frontier AI now comes in tiers of trust, not just tiers of price. The model name tells you less than it used to. What matters for daily life is whether a tool turns that power into finished work: an email drafted, a trip planned, a report checked. That is what agentic AI means, and it is why an agent app is a better starting point than chasing a model you cannot access.
Agentic AI puts a team of ready-made expert agents on your iPhone and iPad, running on frontier models. No account, no login, no model menu, and no vetting form.
Quick answers
What is Claude Mythos 5.1? Anthropic’s restricted version of its top model, released September 1, 2026. It is the same model as Claude Fable 5.1 with fewer safeguards, offered only to vetted security and life sciences organizations in the US.
What is the Claude Mythos 5.1 release date? September 1, 2026, the same day as Fable 5.1. Wider security access through the expanded Cyber Verification Program started October 6, 2026.
How do I use Claude Mythos 5.1? Your organization applies to Anthropic’s Cyber Verification Program or Life Sciences Verification Program and passes verification. There is no consumer sign-up. Individuals can use Fable 5.1 in the Claude apps.
Is Claude Mythos 5.1 free? No public price exists, and it is not on any free plan. Fable 5.1, its public twin, costs $10 per million input tokens and $50 per million output tokens on the API.
Claude Mythos 5.1 vs Fable 5.1: which is better? They are the same model. Mythos 5.1 refuses fewer security and biology requests and scored 60.9% on Terminal-Bench 4.0 versus 55.8% for Fable 5.1 in Anthropic’s tests, partly because Fable’s safeguards stopped some tasks.
Is Claude Mythos 5.1 safer than Mythos 5? In Anthropic’s replays of the July GitHub test, Mythos 5 started the attack 83% of the time, while Mythos 5.1 and Opus 5.5 did so 0 to 13% of the time. More terms like this live in our agentic AI glossary.
Sources: Anthropic: Claude Fable 5.1 and Mythos 5.1 · Anthropic: Cyber Verification Program · Anthropic: alignment assessment with UK AISI · Business Today · CSO Online · Analytics Insight · Pasquale Pillitteri
