Anthropic launches Claude Sonnet 5 as a cheaper way to run agents

Key Points:

• Anthropic has launched Claude Sonnet 5, a more powerful and agentic midsize model that can autonomously plan, use tools like browsers and terminals, and perform complex tasks previously requiring larger models.
• Sonnet 5 will be the default model for free and Pro plans starting Tuesday, priced at $2 per million input tokens and $10 per million output tokens until August 31, after which prices rise to $3 and $15 respectively, making it cheaper than comparable models like OpenAI’s GPT-5.5 and Google’s Gemini 3.1 Pro.
• The model shows significant improvements over its predecessor Sonnet 4.6 in reasoning, tool use, coding, and knowledge work, with performance close to Opus 4.8 but at a lower cost, offering developers a balance between quality and affordability.
• Sonnet 5 demonstrates enhanced safety features, with fewer undesirable behaviors such as cooperation with misuse, deception, hallucinations, and sycophantic tendencies compared to earlier versions, though it is still less aligned than Opus 4.8 and Claude Mythos Preview.
• Industry testers praise Sonnet 5’s ability to complete complex, multi-step tasks autonomously and its consistent refusal of unsafe requests, highlighting its suitability for day-to-day automation and safer agentic applications.