Anthropic has released Claude Sonnet 5, a new agentic AI model designed to perform complex planning, tool use, and autonomous coding tasks at a lower cost than previous Opus-class models. The update narrows the performance gap with Opus 4.8 while offering significant improvements in reasoning, safety, and execution over its predecessor, Sonnet 4.6.
- Performance: Sonnet 5 achieves results close to Opus 4.8 on agentic benchmarks like BrowseComp and OSWorld-Verified, allowing users to balance cost and accuracy via adjustable effort levels.
- Safety: Evaluations show lower rates of hallucination, sycophancy, and undesirable behaviors compared to Sonnet 4.6, with default cyber safeguards enabled due to slightly higher partial success in exploit generation.
- Pricing: Introductory pricing is set at $2 per million input tokens and $10 per million output tokens through August 31, 2026, after which standard rates will be $3 and $15 respectively.
- Availability: The model is available across all plans as the default for Free and Pro users, and accessible via the Claude API, Claude Code, and major cloud platforms.
The release enables developers to deploy powerful agentic workflows for software engineering and automation at a more efficient price point, with partner feedback highlighting its ability to complete multi-step tasks that previously stalled. It serves as a cost-effective alternative to Opus 4.8 for many use cases while maintaining robust safety standards.