What Is Claude Mythos? Anthropic's Next-Gen AI Model
What Is Claude Mythos?
Claude Mythos (internally codenamed Capybara) is Anthropic’s next-generation AI model — accidentally revealed through a data leak in late March 2026. Anthropic confirmed its existence, calling it a “step change” in performance.
Last verified: April 2026
Quick Facts
| Detail | Info |
|---|---|
| Codename | Capybara / Mythos |
| By | Anthropic |
| Status | Early access only (select customers) |
| Leaked | Late March 2026 |
| Capabilities | Multi-step autonomous research, enhanced reasoning |
| Safety concern | Anthropic warns of cyberattack risk |
How It Was Revealed
In late March 2026, a data leak from Anthropic included nearly 3,000 internal assets: details about Claude Mythos, an “exclusive CEO event,” PDFs, images, and other unpublished materials. Fortune broke the story.
Anthropic quickly confirmed the model exists and is being tested, describing it as a “step change” in capabilities compared to Claude Opus 4.6.
What We Know
Capabilities
- Autonomous multi-step tasks — Can execute complex research workflows without human intervention at each step
- Step change over Opus 4.6 — Significant capability improvement (not incremental)
- Enhanced reasoning — Internal benchmarks show major jumps on hard tasks
Safety Concerns
Anthropic is privately warning top government officials that Mythos makes large-scale cyberattacks significantly more likely. From Axios:
“Anthropic is privately warning top government officials that its not-yet-released model — currently branded ‘Mythos’ — makes large-scale cyberattacks much more likely in 2026.”
This is consistent with Anthropic’s Responsible Scaling Policy — they flag dangerous capabilities before wide release.
Availability
- Not publicly available — Only select early access customers have access
- No release date — Anthropic hasn’t announced when (or if) Mythos will become generally available
- Safety testing ongoing — The model is likely undergoing red-teaming and safety evaluation
Claude Model Timeline
| Model | Released | Generation |
|---|---|---|
| Claude 3 (Opus/Sonnet/Haiku) | March 2024 | 3rd gen |
| Claude 3.5 Sonnet | June 2024 | 3.5 gen |
| Claude 4 Opus/Sonnet | 2025 | 4th gen |
| Claude 4.5 Sonnet/Opus | Late 2025 | 4.5 gen |
| Claude Opus 4.6 / Sonnet 4.6 | Feb 2026 | 4.6 gen |
| Claude Mythos | TBD | Next gen |
What to Expect
If Mythos follows Anthropic’s pattern:
- Extended safety testing before release
- API access first for developers
- Claude.ai integration shortly after
- Pricing premium — likely more expensive than Opus 4.6
The model represents Anthropic’s bid to maintain frontier-model leadership against GPT-5.4 and Gemini 3.1 Pro.
Last verified: April 2026