systemcards.org

Claude Mythos

Publisher
anthropic first-party
Type
access_policy
Models
Claude Mythos 5, Claude Mythos 5.1, Claude Mythos 5.2
Published
unknown
Status
active
Safety evals
none found
Openness
restricted access
Source
https://www.anthropic.com/claude/mythos
Related
other https://claude.com/form/mythos-access-interest
announcement https://www.anthropic.com/claude-fable-and-mythos-5-1
First seen
2026-09-16T13:06:47Z
Versions
6

Report an issue with this entry

Version history

FetchedContent typeSizeFingerprintOriginalWhat changed
2026-11-10T06:18:19Z text/html 173,099 54fd43075941845f archived copy Page retitled from Claude Mythos 5.1 to Claude Mythos 5.2. Mythos 5.2 is available to vetted cyberdefenders and life scientists through the Cyber and Life Sciences Verification Programs. Pricing drops to 8/40 USD per million tokens, and fast mode (16/80) is limited to trusted-access programs. Eligible Cyber Verification Program customers with ZDR may keep it until Enterprise Frontier Safeguards is available. A Fable 5.2 benchmark table was added.
2026-11-03T06:21:33Z text/html 158,155 123145fa9c3c59d7 archived copy —
2026-10-21T06:21:50Z text/html 195,196 3be3285e07b065d7 archived copy Extractor drift: 5 line(s) missing from the extracted text are still present in the raw capture; not a publisher deletion.
2026-10-08T06:21:16Z text/html 275,098 b9a4f8fc2c3f024b archived copy —
2026-09-29T06:21:31Z text/html 288,244 9cc0b7f9a1581c21 archived copy —
2026-09-16T13:06:47Z text/html 290,986 4b602353ba523df9 archived copy initial version

Provenance

Show 9 changelog entries (when, how, and why this entry changed)
  1. 2026-09-16T13:06:47Z add run 2026-09-16T12:52Z-local · agent · lead: agent_search
    Anthropic's standing access-tier overview page for Claude Mythos, which I fetched and read. It states that Claude Mythos 5.1 is available only to vetted cyberdefenders and life scientists through Anthropic's trusted access programmes, and it is the page that documents both gates: the Life Sciences Verification Program (invite-only beta, reduced biology safeguards, currently US organizations only) and the Cyber Verification Program (reduced cyber safeguards for defensive security work, with Mythos access described as coming soon). It also records the default 30-day data-retention requirement for safety monitoring and the relationship to the generally-available Claude Fable 5.1 (same underlying model, different safeguards). This is an access-tier overview naming specific gated models, which TASK.md lists explicitly under doc_type access_policy, and it is distinct from the dated Mythos posts already catalogued (expanding-support-for-scientists, bringing-claude-mythos-5-to-more-defenders, the Fable/Mythos 5.1 launch post). It is an evergreen page carrying no publication date of its own, so publication_date is null rather than guessed; the database already holds several undated access-policy rows of this kind. Openness is restricted because every model named is gated to vetted parties.
    Criteria: document_retrievable=pass; min_publication_date=FAIL; publisher_on_allowlist=pass · attested about_a_specific_model_or_eval=true · attested covered_model_class=true · attested distinct_model_release=true · attested notable_release=true · attested primary_source=true
  2. 2026-09-29T06:21:31Z new_version run 2026-09-29T06:19Z-local · monitor · lead: monitor
    Content fingerprint changed on scheduled re-fetch
  3. 2026-10-08T06:21:16Z new_version run 2026-10-08T06:19Z-local · monitor · lead: monitor
    Content fingerprint changed on scheduled re-fetch
  4. 2026-10-21T06:21:50Z new_version run 2026-10-21T06:19Z-local · monitor · lead: monitor
    Content fingerprint changed on scheduled re-fetch
  5. 2026-10-21T06:21:50Z field_update run 2026-10-21T06:19Z-local · monitor
    change_summary: null → "Extractor drift: 5 line(s) missing from the extracted text are still present in the raw capture; not a publisher deletion."
    Deleted lines found verbatim in the new raw capture
  6. 2026-11-03T06:21:33Z new_version run 2026-11-03T06:16Z-local · monitor · lead: monitor
    Content fingerprint changed on scheduled re-fetch
  7. 2026-11-10T06:18:19Z new_version run 2026-11-10T06:16Z-local · monitor · lead: monitor
    Content fingerprint changed on scheduled re-fetch
  8. 2026-11-10T06:23:13Z field_update run 2026-11-10T06:16Z-local · agent · lead: manual
    change_summary: null → "Page retitled from Claude Mythos 5.1 to Claude Mythos 5.2. Mythos 5.2 is available to vetted cyberdefenders and life scientists through the Cyber and Life Sciences Verification Programs. Pricing drops to 8/40 USD per million tokens, and fast mode (16/80) is limited to trusted-access programs. Eligible Cyber Verification Program customers with ZDR may keep it until Enterprise Frontier Safeguards is available. A Fable 5.2 benchmark table was added."
    Diff v1049 to v1116 is a substantive update of the Mythos program page for the Mythos 5.2 launch on 2026-11-09: the gated model, pricing, fast-mode gating and the data-retention policy all changed.
  9. 2026-11-10T06:23:16Z field_update run 2026-11-10T06:16Z-local · agent · lead: monitor
    model_names: ["Claude Mythos 5", "Claude Mythos 5.1"] → ["Claude Mythos 5", "Claude Mythos 5.1", "Claude Mythos 5.2"]
    The evergreen Mythos program page (anthropic.com/claude/mythos) now names Claude Mythos 5.2 as the gated model (version 1116, fetched 2026-11-10) and lists Mythos 5.2 in the Cyber Verification Program scope. Adding it keeps the restricted-access filter surfacing this program page for the new model.

Extracted text

Show extracted text (from the latest fetched version; formatting is lossy — use the original for reading)
Claude Mythos 5.2
Our most capable model class for cybersecurity and biology research.
Announcements
- Introducing Claude Mythos 5.2Claude Mythos 5.2 is our most capable model for cybersecurity and life sciences research. It’s available to vetted organizations through our verification programs. View more
- Expanding the Cyber Verification ProgramSecurity teams can now apply for advanced cyber capabilities on Claude Opus 5.5, Claude Sonnet 5.5, Claude Mythos 5.1, and future models. View more
- Introducing Claude Mythos 5.1Claude Mythos 5.1 is our newest Mythos-class model, with gains in cybersecurity and biology. Access remains limited to a small set of vetted organizations. View more
- Claude Mythos 5 export controls have been liftedWe have restored access to Mythos 5 for a set of US organizations, following the US government’s approval. View more
- Claude Mythos 5 is currently unavailableWe apologize for this disruption to our customers and are working to restore access as soon as possible. View more
- Claude Mythos 5Claude Mythos 5 is the latest update for Mythos Preview, seeing gains in cybersecurity, biology, and healthcare benchmarks. Given its capabilities, Mythos 5 is currently only available to a small group of vetted partners with a goal of opening up more broadly in the future. View more
- Expanding Project GlasswingWe’re extending Project Glasswing to approximately 150 new organizations in more than fifteen countries. View more
Availability and pricing
Claude Mythos 5.2 is available to vetted cyberdefenders and life scientists through our trusted access programs. Life sciences researchers can apply for advanced capabilities through the Life Sciences Verification Program. Security teams can apply for advanced cyber capabilities on Claude Mythos 5.2, Claude Opus 5.5, Claude Sonnet 5.5, Claude Mythos 5.1, and future models through the Cyber Verification Program.
Pricing for Claude Mythos 5.2 starts at $8 per million input tokens and $40 per million output tokens.
Fast mode for Claude Mythos 5.2 is also available through our trusted access programs. It costs $16 per million input tokens and $80 per million output tokens.
Claude Security also now runs on Mythos 5.2.
Safeguards
Because Mythos 5.2 is highly capable for cybersecurity and biology research, it could be used both for good and for harm. We currently provide access to a set of vetted organizations through our verification programs.
We extensively test and evaluate our models to ensure they meet Anthropic’s standards for safety, security, and reliability. The accompanying system card covers safety results in depth.
Data retention
Using Mythos models requires 30-day data retention for safety monitoring by default. With Enterprise Frontier Safeguards (EFS), automated safety monitoring is run on prompts and responses that are stored in a customer’s infrastructure, making it compatible with zero data retention (ZDR). Until EFS is available, eligible Cyber Verification Program customers who already have ZDR can keep it for Mythos 5.2.
Learn more about data retention. Learn more about Enterprise Frontier Safeguards.
Benchmarks
Claude Fable 5.2 and Claude Mythos 5.2 are the same underlying model. Results below are for Fable 5.2 unless noted; where Mythos 5.2 results are shown, the gap reflects tasks where Fable 5.2’s safeguards intervene.
|  | Fable 5.2 | Fable 5.1 | Opus 5.5 | GPT-6 Astra | Gemini 4 Carbon |
|---|---|---|---|---|---|
| Agentic codingFrontierCode 2.0 | 49.1%Xhigh effort | 32.4%Xhigh effort | 45.5% | 42.8% | 41.1%Medium effort |
| Agentic codingFrontierSWE v2 | 69.9% | 56.3% | 62.3% | 65.5% | 61.2% |
| Business workflowsAutomationBench | 51.4% | 31.4% | 42.5% | 41.4% | 53.9% |
| Agentic scientific researchTerminal-Bench-Science 0.1 | 74.6%75.3% (Mythos 5.2) | 52.6% | 58.7% | 68.1% | 64.8% |
| CybersecurityCyScenarioBench† | 76.3%Mythos 5.2 | 61.7%Mythos 5.1 | 67.6%without safeguards | 61%GPT-6 Cyber | — |
| CybersecurityCVE Rediscovery | 89.6%Mythos 5.2 | 78.3%Mythos 5.1 | 81.3%without safeguards | — | — |
| Protein designDe novo binder design (24 hr) | 85.1%Mythos 5.2 | 79.1%Mythos 5.1 | 82.6%without safeguards | — | — |
| Agentic searchWANDR (relaxed score) | 82.1% | 68.7% | 72.3% | — | — |
| Computer useOSWorld 2.1 (offline set, partial) | 89.3% | 83.5% | 87.2% | 79.5% | 81.5% |
| RoboticsRoboDojo Sim‡ | 64.9% | 27.6% | 39.4% | 29.8% | — |
| Knowledge workAA-Briefcase v1.1 | 66.1%rubric | 61.5%rubric | 61.2%rubric | 52.0%rubric | 61.8%rubric |
|  | 1785Elo | 1669Elo | 1800Elo | 1569Elo | 1611Elo |
| Knowledge workRemote Labor Index (RLI) | 41.3%High effort | 17.9% | 21.3% | 20.8% | — |
| Multidisciplinary reasoningHumanity’s Last Exam (with tools) | 74.2% | 65.6% | 67.7% | 57.2% | — |
| Visual chart recognitionChartography (no tools) | 73.4% | 44.8% | 64.4% | 71.0% | 72.5% |
For details on evaluation methodology, see the Claude Fable 5.2 and Claude Mythos 5.2 System Card. Scores for Fable 5.2, Fable 5.1, and Opus 5.5 show results with safeguards on unless otherwise noted. All scores show each model’s highest-available effort unless otherwise noted. † On CyScenarioBench, each lab’s scores are based on unique tasks. ‡ RoboDojo Sim shows the highest-available effort and was run internally.
Frequently asked questions
Making models as capable as Mythos Preview generally available comes with risks. The model’s capabilities in specific areas like cybersecurity and biology are advanced enough that they could be misused to create cyberattacks or dangerous weapons. To release Mythos-level capabilities more broadly, we developed Claude Fable models with safeguards that block or limit its performance in risky areas.
Security teams can apply for advanced cyber capabilities on Claude Mythos 5.2, Claude Opus 5.5, Claude Sonnet 5.5, Claude Mythos 5.1, and future models through the Cyber Verification Program. Life sciences researchers can apply for advanced capabilities through the Life Sciences Verification Program.