Cheaper but safer? Claude Sonnet 5 bets that agentic AI can scale without spiraling costs or cyber risks

Cheaper but safer? Claude Sonnet 5 bets that agentic AI can scale without spiraling costs or cyber risks
Anthropic’s new Claude Sonnet 5 lands at a fraught moment for AI agents, promising long‑running autonomous work for the masses while regulators, rivals, and customers all focus on cost and safety.

Anthropic’s pitch: more autonomy, less risk

On June 30, Anthropic unveiled Sonnet 5 as “the most agentic Sonnet model yet,” built to make plans, use browsers and terminals, and run autonomously at a level that recently required larger, pricier systems. The company stresses that Sonnet 5’s performance is “close to that of Opus 4.8, but at lower prices,” and that safety tests show “a much lower ability to perform cybersecurity tasks than our current Opus models.”

Axios frames the release as a deliberate move to bring agentic capabilities “to everyday users without the same cyber-risk profile as its most powerful systems,” noting that Sonnet 5 wasn’t deliberately trained on cybersecurity tasks and has a “much lower ability” to perform dangerous cyber activities than Opus.

Pricing pressure and the agent race

TechCrunch situates Sonnet 5 in a broader shift where “agentic capabilities becomes table stakes,” arguing the real battle is now “how cheaply they can do it and how reliably without human oversight.” The model “promises performance close to that of Opus 4.8, but for much lower costs,” and is cheaper than both OpenAI’s GPT‑5.5 and Google’s Gemini 3.1 Pro at launch prices.

The Next Web underscores the cost angle, calling Sonnet 5 “its most agentic mid-tier model yet” that “runs close to the flagship Opus 4.8 on many tasks, but costs less than half as much,” with introductory pricing of $2 per million input tokens and $10 per million output tokens through August 31, before rising to $3 and $15.

Deployment and ecosystem response

Anthropic says Sonnet 5 is available across all plans and is the default for Free and Pro tiers, as well as on the Claude API and Claude Code. Axios adds that this comes as Anthropic remains in talks with the Trump administration after it was asked to pull its most powerful models, Mythos and Fable, over security concerns, with those systems returning only on a restricted basis.

On the tools side, Perplexity AI CEO Aravind Srinivas highlighted Sonnet 5’s rapid integration, calling it a “New orchestrator model for Computer users (Pro and Max): Sonnet 5,” and amplifying Perplexity’s announcement that “Claude Sonnet 5 is now available in Perplexity for Pro and Max subscribers” and selectable as an orchestrator model.

Across human analysts and Anthropic’s own framing, Sonnet 5 is cast as a compromise candidate: near‑flagship autonomy aimed at enterprise and everyday agents, but priced and constrained to avoid the financial and cyber risks that have begun to shadow top‑end AI.

Continue reading https://foxvector.com/stories/019f604b-733d-17b4-7395-03cea8d267f6