Claude Capybara — Everything You Need to Know

Independent research hub covering the Capybara tier — Anthropic's most powerful level of AI models, above Opus. Benchmarks, capabilities, cybersecurity implications, release timeline, and industry impact.

What We Cover

Eight research clusters spanning the full Capybara story.

Core Information

What the Capybara tier is, how it fits into Anthropic's hierarchy (Haiku → Sonnet → Opus → Capybara), and what makes it different.

Read the full guide →

Benchmarks and Performance

Terminal-Bench 2.0, SWE-bench, ARC-AGI-2, GPQA Diamond — scores, context, and real-world implications.

See benchmark data →

Cybersecurity

The Chinese espionage campaign, Accenture Cyber.AI, stock market impact, and dual-use AI in offense and defense.

Read about cybersecurity →

Model Comparisons

Claude Capybara vs GPT-5, Gemini, and previous Claude models — real benchmark data, not marketing.

Compare models →

Safety and Ethics

ASL-4 protocols, the danger debate, responsible scaling, and governance implications.

Explore safety →

Release and Pricing

Timeline predictions from Polymarket, API availability, and pricing estimates.

Check timeline →

The Leak Story

How ~3,000 CMS files went public and what they revealed about the Capybara tier.

Read the story →

Developer Resources

Coding benchmarks, agent workflows, API guides, and practical developer information.

Developer guide →

All Articles

Core Information

• What Is Claude Capybara?
• Claude Capybara Tier Explained
• Claude Model Tiers: Haiku, Sonnet, Opus, Capybara

Capabilities and Benchmarks

• Claude Capybara Benchmark Results
• Claude Capybara Capabilities
• Claude Capybara for Coding

Cybersecurity

• Claude Capybara Cybersecurity
• AI Cybersecurity Risk 2026

Comparisons

• Claude Capybara vs GPT-5
• Claude Capybara vs Opus
• Claude Capybara vs Gemini

Safety and Ethics

• Is Claude Capybara Dangerous?
• Claude Capybara ASL-4 Safety

Access and Pricing

• Claude Capybara Release Date
• Claude Capybara Pricing
• Claude Capybara API Access
• How to Get Claude Capybara

The Leak and Business

• Anthropic Capybara Leak

Developer

• Claude Capybara for Developers
• Claude Capybara Agent Workflows

Questions About Claude Capybara

What is the Capybara tier?

Capybara is Anthropic's newest model tier, above Opus. The hierarchy is Haiku, Sonnet, Opus, Capybara. Claude Mythos is the first Capybara-tier model — leaked documents describe it as "a step change" in AI capabilities.

When will it be available?

No confirmed date. Polymarket shows 27% chance by April 30 and 45% by June 30, 2026. Phased rollout expected, starting with cyber defense organizations.

How does it compare to GPT-5?

Closely matched on many benchmarks. Opus 4.6 ties GPT-5.4 on Terminal-Bench (65.4%), but GPT leads SWE-Bench Pro coding (57.7% vs 45.9%) while Opus leads GPQA Diamond reasoning by 3.5 points. Capybara is expected to surpass both.

Is this an official Anthropic site?

No. claudecapybara.pro is fully independent. Not affiliated with Anthropic. Visit anthropic.com for official information.

Is Claude Capybara dangerous?

It raises dual-use concerns in cybersecurity. The September 2025 espionage campaign showed current Claude models can be weaponized. ASL-4 safety protocols aim to mitigate these risks.