Claude Mythos (Glasswing) is a Restricted cyber from Anthropic (United States).
One-line fit: best used for Defensive cyber partners.
Key specifications
| Provider | Anthropic |
|---|---|
| Model | Claude Mythos (Glasswing) |
| Type | Restricted cyber |
| Context window | Gated |
| Pricing (indicative) | Partner only |
| Company category | Frontier Lab |
| Region | United States |
| Official site | https://www.anthropic.com |
Benchmarks and public standing
Not public leaderboard
Treat scores as directional. SWE-bench, Terminal-Bench, LMSYS Arena, Artificial Analysis, and vendor cards use different harnesses and are not always comparable 1:1. Re-check the latest official model card before decisions.
What this model is good at
Defensive cyber partners
This aligns with Anthropic’s broader strengths:
- Hard software engineering
- Long-form writing and research
- Computer use / Claude Code agents
- Enterprise trust and safety brand
Limitations and watch-outs
Not generally available
- Rate limits, regional availability, and data retention policies vary by plan.
- Agent harness quality (Cursor, Claude Code, Codex, custom tools) can change outcomes more than raw model Elo.
- Open-weight availability (if any) is separate from hosted API quality and safety filters.
Ideal users
- Product engineers shipping features that match: Defensive cyber partners
- Teams standardizing on the Anthropic ecosystem
- Agent builders who need a Restricted cyber
When to pick something else
- Cheaper volume: compare lower tiers from the same lab or open Chinese/EU alternatives.
- Maximum hard SWE thoughtfulness: compare Claude Fable-class and other coding flagships.
- Giant multimodal corpora / Workspace: compare Gemini-class models.
- Self-host / open weights: Llama, Qwen, DeepSeek, GLM, Mistral open lines.
Parent company
Anthropic – Safety-focused frontier lab. Claude models are widely preferred for thoughtful coding, long writing, and high-stakes knowledge work.
Related models from Anthropic
- Claude Fable 5 – Mythos-class flagship
- Claude Opus 4.x – Prior flagship tier
- Claude Sonnet 5 – Default Pro/Free workhorse
- Claude Haiku – Fast/cheap
Data is curated for AIForumSphere model directory mid-2026 directional figures.
