Anthropic has introduced Claude Sonnet 5, a midsize AI model designed to perform complex agentic tasks with enhanced capabilities and at significantly lower cost than rivals like OpenAI’s GPT-5.5 and Google’s Gemini Pro.
- Improved agentic capabilities including autonomous tool use and reasoning
- Lower priced than competing AI agents through August 31
- Enhanced safety features reduce undesirable behaviors and misuse
What happened
Anthropic has launched Claude Sonnet 5, a midsize AI model that significantly upgrades its agentic abilities. The model can autonomously plan and execute complex multi-step workflows while utilizing tools such as browsers and terminals. Anthropic positions Sonnet 5 as capable of performing tasks previously requiring much larger and more expensive models.
Claude Sonnet 5 is immediately available as the default model for both free and Pro subscription plans. Initial pricing through August 31 is set at $2 per million input tokens and $10 per million output tokens, undercutting popular alternatives including OpenAI’s GPT-5.5 and the earlier Opus 4.8 model. After this introductory period, input token pricing will rise to $3 per million, still remaining competitive.
Why it matters
As agentic capabilities become fundamental across major AI platforms, Sonnet 5’s introduction demonstrates a shift in market focus toward delivering these functions more affordably and safely. While leading models like OpenAI's GPT-5.6 and Google’s Gemini 3.5 emphasize agentic planning, Sonnet 5 balances strong performance with cost efficiency, expanding access to developers.
Performance benchmarks indicate Sonnet 5 delivers substantial improvements over its predecessor Claude Sonnet 4.6, especially in reasoning, coding, and autonomous task completion. Its agentic coding proficiency reaches 63.2%, closing the gap with Opus 4.8’s 69.2%, while surpassing Opus on knowledge work challenges. Additionally, Sonnet 5 improves safe usage by reducing hallucinations, refusal of malicious requests, and resilience to prompt injection attacks, though more advanced Anthropic models retain the lead in safety.
What to watch next
Developers and enterprises will assess how Sonnet 5 performs in real-world automation and agentic workflows, particularly given its competitive pricing and improved reliability. The model’s ability to complete complex multi-step jobs autonomously, as demonstrated in early tests, is likely to attract widespread use in productivity and software automation.
Anthropic’s pricing strategy through August 31 will be critical to driving adoption momentum. Observers will watch whether Sonnet 5’s balance of affordability, capability, and safety influences broader AI pricing trends and user expectations for midsize agentic models. Continued safety enhancements and feature upgrades will also be key to maintaining its competitive positioning against upcoming releases from OpenAI and Google.