Claude Sonnet 5 ships; performance near Opus 4.8 at lower cost

Claude Sonnet 5 is now available, representing a substantial improvement over Sonnet 4.6 with enhanced agentic capabilities. The model is designed to be the most agentic Sonnet model yet, capable of making plans, using tools like browsers and terminals, and running autonomously.

Performance:

• Sonnet 5 performance is close to Opus 4.8 but at lower prices
• Substantial improvements over Sonnet 4.6 in reasoning, tool use, coding, and knowledge work
• Narrower gap between Sonnet and Opus models in agentic capabilities

Availability:

• Available across all plans: default for Free and Pro, available to Max, Team, and Enterprise users
• Available in Claude Code and on Claude Platform
• Introductory pricing: $2 per million input tokens and $10 per million output tokens through August 31, 2026
• Standard pricing after August 31: $3 per million input tokens and $15 per million output tokens
• API access via claude-sonnet-5

Safety:

• Overall lower rate of undesirable behaviors than Sonnet 4.6
• Generally safer to use in agentic contexts
• Much lower ability to perform cybersecurity tasks compared to Opus models
• Launched with cyber safeguards enabled by default (same as Opus 4.7 and 4.8, but less strict than Fable 5)
• Cannot successfully develop working exploits for software vulnerabilities

Rate Limits: Rate limits have been increased across Chat, Cowork, Claude Code, and Claude Platform to accommodate higher token usage of higher effort levels.