Claude Opus 4
v20250514Anthropic
RETIRED: Anthropic retired Claude Opus 4 (claude-opus-4-20250514) on 2026-06-15 (deprecated 2026-04-14); API requests now fail. Recommended replacement: Claude Opus 4.8. Historically Anthropic's most powerful model of May 2025, with exceptional reasoning, coding (72.5-79.4% SWE-bench in high-compute), and agentic capabilities.
Trust Vector Analysis
Dimension Breakdown
🚀Performance & Reliability+
Historical evaluation: best-in-class performance at release (79.4% SWE-bench high-compute, 90% AIME high-compute). Model retired 2026-06-15 and is no longer served.
Industry-standard coding benchmarks
PhD-level reasoning benchmarks
Comprehensive knowledge testing
Internal testing
Median latency
Official specification
🛡️Security+
Flagship security with ASL-3 standard and Constitutional AI. Strongest safety guardrails.
OWASP LLM01 testing
Adversarial prompt testing
Privacy policy analysis
Comprehensive safety testing
Security features review
🔒Privacy & Compliance+
Exceptional privacy. Ephemeral data handling, HIPAA eligible, strongest compliance for regulated industries.
Enterprise documentation review
Policy analysis
Terms review
Data protection review
Certification verification
Data handling review
👁️Trust & Transparency+
Excellent transparency with extended thinking and comprehensive system card. Best-in-class guardrails.
Reasoning transparency evaluation
Factual QA testing
Bias benchmark evaluation
Qualitative assessment
Documentation review
Public disclosure review
Safety mechanism analysis
⚙️Operational Excellence+
Model retired 2026-06-15 on Anthropic-operated platforms; API requests fail. Migration target is Claude Opus 4.8. Versioning, ecosystem, and overall scores reduced to reflect retirement.
API design review
SDK quality review
Versioning policy review
Monitoring tools review
Support assessment
Ecosystem analysis
License review
- +Highest performance: 79.4% SWE-bench in high-compute (best overall)
- +90% AIME math in high-compute mode (exceptional reasoning)
- +Extended thinking for complex multi-step reasoning
- +Strongest privacy: ephemeral data, HIPAA eligible, ASL-3 security
- +200K context window for large documents
- +Best-in-class Constitutional AI safety guardrails
- !RETIRED 2026-06-15 — no longer available on the Claude API; requests fail (migrate to Claude Opus 4.8)
- !Premium pricing ($15/$75 per 1M tokens)
- !Higher latency (~2.5s p50, 5.5s p95)
- !Training cutoff March 2025
- !Overkill for simple tasks (cost and latency)
- !32K max output vs 64K for Sonnet 4
Use Case Ratings
code generation
Historically best-in-class coding (79.4% SWE-bench high-compute). Retired — use Opus 4.8.
customer support
Excellent but potentially over-powered and expensive for standard customer support.
content creation
Exceptional creative writing with nuanced understanding and natural style.
data analysis
Superior analytical capabilities with extended thinking for complex analysis.
research assistant
Outstanding for research. Extended thinking enables deep analysis. 200K context for long documents.
legal compliance
Best for legal work in its era. HIPAA eligible, ephemeral data, ASL-3 security. Careful reasoning.
healthcare
Flagship for healthcare in its era. HIPAA eligible, strongest privacy, careful medical reasoning.
financial analysis
Exceptional for complex financial modeling and analysis. 90% AIME math in high-compute.
education
Excellent for education with patient, detailed explanations and strong knowledge base.
creative writing
Outstanding creative capabilities with nuanced character development and storytelling.