Claude Mythos 5.2
Released Nov 9, 2026 · knowledge cutoff 2026-06
- Status
- Frontier
- Location
- United States
- Type
- LLM
- Modality
- Multimodal
- Context window
- 1M
- Max output
- 128K
- Speed
- —
- Price ($/MTok in / out)
- $8 / $40
- Cost per task
- —
- Open weights
- No
News1 story
Strengths & weaknesses
Strengths
- Rediscovered 89.6% of known CVEs in Anthropic's evaluation, up from Mythos 5.1's 78.3%
- Completed 76.3% of CyScenarioBench challenges, the strongest result Anthropic reports for any model
- Top score on Anthropic's de novo protein binder design task at 85.1%
- Same capabilities and $8/$40 pricing as Fable 5.2 without Fable's cyber and biology safeguards
Weaknesses
- Only for organisations verified through Anthropic's Cyber or Life Sciences Verification Programs, plus Claude Enterprise via Claude Security
- Refused 82.8% of malicious requests in the system card's misuse test, down from Mythos 5.1's 88.5%
Same model as Claude Fable 5.2 with fewer safeguards. Anthropic's model page (platform.claude.com/docs/en/models/mythos-5-2/overview) says it is available only to organisations verified through its verification programs, with the same specs and pricing as Fable 5.2 (1M context, 128K output, $8/$40, cache reads $0.20, June 2026 knowledge cutoff); the launch post adds that Claude Enterprise customers can use it through Claude Security without verification, and Mythos fast mode is limited to verified organisations. Ids: claude-mythos-5-2 (Claude API, Google Cloud, Foundry), anthropic.claude-mythos-5-2 (Bedrock). Benchmarks stay null: the system card reports Mythos 5.2 only on untracked suites (CVE Rediscovery 89.6%, CyScenarioBench 76.3%, De Novo Binder Design 85.1%, all without safeguards) and gives Fable 5.2's tracked scores under Fable's name, so they aren't copied across. Retirement is 'not sooner than November 9, 2027', a floor. predecessorId is null: Anthropic gives a migration path from Mythos 5.1 but no replacement statement, and Mythos 5.1 isn't tracked. The launch table gives Mythos 5.2 75.3% on Terminal-Bench-Science 0.1 (Fable 5.2 74.6%), also untracked.
Benchmarks0 of 10 reported
MMLU-Proretired
GPQA Diamond
SWE-bench Verified
SWE-bench Pro
Terminal-Bench 2.1
AIME (latest)retired
Humanity's Last Exam
LMArena Elo
GDPval-AA v2
ARC-AGI-2
For developers
API model strings
- anthropic
claude-mythos-5-2 - aws-bedrock
anthropic.claude-mythos-5-2 - google-vertex
claude-mythos-5-2
Licence
Proprietary — weights not released
Retirement
No retirement announced
Lineage
No recorded predecessor