Claude Sonnet 5: 4 Essential Things You Need to Know
Get the pricing, performance, tool use, and safety details that matter for creators and developers.
Anthropic unveiled Claude Sonnet 5, a midsize model that promises stronger agentic capabilities at a lower price point. In this article you’ll learn four concrete facts about the new model—how much it costs, how it outperforms its predecessor, what tools it can use autonomously, and how Anthropic has tightened its safety guardrails.
Claude Sonnet 5 Pricing and Availability
Anthropic is positioning Sonnet 5 as the default model for both free and Pro plans. The company announced a launch price of $2 per million input tokens and $10 per million output tokens through August 31, after which the rates rise to $3 and $15 respectively. This makes Sonnet 5 cheaper than the larger Opus 4.8 model while still delivering near Opus quality on many tasks.
02Agentic Performance Improvements in Claude Sonnet 5
Sonnet 5’s biggest claim is its boost in agentic coding. On a benchmark that measures how well a model can write and run code without human prompting, Sonnet 5 scored 63.2 % – a jump from Sonnet 4.6’s 58.1 % and only a few points shy of Opus 4.8’s 69.2 %. The model also edged out its predecessor on knowledge work tests, slightly surpassing Opus 4.8 on a complex reasoning benchmark.
Tool Use and Autonomous Tasks with Claude Sonnet 5
Anthropic says Sonnet 5 can plan, browse the web, and run terminal commands on its own. In a real world test, a senior engineer at Zapier asked the model to update Salesforce account tiers and then email a launch announcement to enterprise contacts. The model completed the entire two part workflow without stopping, something earlier versions often left unfinished. This level of autonomy, previously reserved for larger, more expensive models, opens the door for everyday developers to build richer automations.
04Safety Guardrails and Misuse Prevention in Claude Sonnet 5
Safety remains a priority. Sonnet 5 shows a lower rate of “undesirable behaviors” such as cooperating with malicious requests or falling for prompt injection attacks, according to Anthropic’s internal evaluations. While it still isn’t as resistant as the Mythos class models, the new guardrails mean the model is more likely to refuse unsafe queries outright, giving developers confidence when deploying it in production pipelines.
Sources: Anthropic Announcement, Claude Pricing Platform, MacRumors, Marktechpost
Claude Sonnet 5’s most compelling feature is its agentic ability to finish multi step tasks at a price that small teams can afford. If you need a model that can run tools and still stay within a tight budget, Sonnet 5 is the clear choice—just keep an eye on its evolving safety filters.