Model Details
Claude Mythos 5.1
Quick answer: Claude Mythos 5.1 is identical to Claude Fable 5.1 at the model level, but ships with more permissive cybersecurity and life-sciences safeguards for vetted individuals and organizations. It's available only through Anthropic's Cyber Verification Program (CVP) and Life Sciences Verification Program (LSVP), announced September 3, 2026. On Terminal-Bench 4.0 (with cybersecurity safeguards relaxed), Mythos 5.1 scores 60.9% — the highest of any model in Anthropic's comparison.
Access Status
Claude Mythos 5.1 is not generally available. Access requires enrollment in one of two trusted access programs:
- Cyber Verification Program (CVP) — reduced cybersecurity safeguards for defensive security work; Mythos-class model access is rolling out to CVP in the near future.
- Life Sciences Verification Program (LSVP) — safeguards designed for professional life-sciences research and development, developed in partnership with the US government; first participants enrolled at launch.
Claude Security, Anthropic's codebase vulnerability-scanning product, is now powered by Claude Mythos 5.1.
At a Glance
Where Claude Mythos 5.1 leads
- 60.9% Terminal-Bench 4.0 — the top score among all models in Anthropic's Fable 5.1/Mythos 5.1 comparison, ahead of Fable 5.1's own 55.8%
- Strongest cyber capabilities of any Anthropic model released to date (with cybersecurity safeguards off), while still falling within the lower risk category of Anthropic's Frontier Compliance Framework
- Improved molecular design capability: assisted binder designs with binding affinities reported as ~10x higher than the best entries in Adaptyv Bio's protein design competitions, across three validated targets
- Better-aligned than its predecessor Mythos 5 on Anthropic's automated behavioral audit (less likely to seek unauthorized resources, use motivated reasoning, or ignore explicit constraints)
Where it lags
- No general availability — gated entirely behind CVP/LSVP enrollment
- Chemical/biological-risk evaluations show increased capability versus Mythos 5, so Anthropic keeps the same restrictive biology safeguards applied to Mythos 5
- Regional availability currently limited to a set of US organizations, expanding via coordination with the US government
Best for: Vetted cyberdefenders and life-sciences professionals operating inside Anthropic's trusted access programs; not applicable for general-purpose deployment.
What Claude Mythos 5.1 Is
Claude Mythos 5.1 is the higher-permission counterpart to Claude Fable 5.1, announced together on September 3, 2026. Anthropic describes both as the same model, differing only in safeguard configuration: Fable 5.1 ships Anthropic's standard production safeguards for general availability, while Mythos 5.1 relaxes specific cybersecurity and life-sciences restrictions for vetted users whose legitimate work is otherwise blocked by those safeguards.
Ahead of release, Anthropic subjected the model to extended red-teaming and evaluation for chemical/biological risk (including a tabletop exercise pairing PhD-level biologists with AI red-teamers) and cyber risk (including external stress-testing by two organizations and Gray Swan). Anthropic reports Mythos 5.1's capabilities exceed Mythos 5's in both areas, but still fall short of the next Responsible Scaling Policy risk tier — so Mythos 5.1 retains the same restrictive biology safeguards used for Mythos 5, and is deployed under Anthropic's Frontier Compliance Framework for cyber risk.
Specifications
| Field | Value |
|---|---|
| Organization | Anthropic |
| Parameters | Undisclosed |
| Context window | Undisclosed |
| Architecture | Undisclosed (Mythos line — same base model as Fable 5.1) |
| License | Proprietary (API only) |
| Modality | Text + Vision (multimodal) |
| Knowledge cutoff | October 2025 (estimated) |
| Access status | Trusted access only (CVP / LSVP), as of Sept 2026 |
Pricing
Pricing for Claude Mythos 5.1 is not publicly disclosed — access is granted per-organization through CVP/LSVP enrollment rather than standard API self-service.
Benchmark Results
Anthropic's only directly reported Mythos 5.1 score in its Fable 5.1/Mythos 5.1 launch comparison is on Terminal-Bench 4.0:
| Category | Benchmark | Mythos 5.1 | Fable 5.1 | Fable 5 | Opus 5 | GPT-5.6 Sol |
|---|---|---|---|---|---|---|
| Agentic coding | Terminal-Bench 4.0 | 60.9% | 55.8% | 42.0% | 52.3% | 37.3% |
Mythos 5.1's remaining capabilities (molecular design, GPU-kernel optimization for computational biology, inference speedups on open-source genomics/protein models) are described qualitatively in Anthropic's release rather than as standard leaderboard scores.
Score reported by Anthropic, September 2026.