Anthropic unveiled Claude Sonnet 5 on Tuesday, positioning the midsize model as the newest baseline for agentic AI work. The firm says Sonnet 5 can make plans, operate browsers and terminals, and execute tasks autonomously—functions that a few months ago required larger, more expensive models.
Pricing reflects that shift. For the remainder of August, the model costs $2 per million input tokens and $10 per million output tokens, after which input costs rise to $3 while output pricing stays the same. Those rates undercut Anthropic’s own Opus 4.8 and sit below the rates of OpenAI’s GPT‑5.5 and Google’s Gemini 3.1 Pro, though Gemini 3.5 Flash remains cheaper.
Performance benchmarks show Sonnet 5 closing the gap with Opus 4.8. On an agentic coding test, Sonnet 5 scored 63.2%, outpacing its predecessor Sonnet 4.6’s 58.1% and approaching Opus 4.8’s 69.2%. In a knowledge‑work benchmark, the new model even edged out Opus 4.8, which has long been praised for deep‑research tasks.
Developers who have already tried the model report noticeable gains in task completion. Daniel Shepard, a senior engineer at Zapier, described a two‑part workflow—updating Salesforce account tiers and sending a launch announcement—that Sonnet 5 completed end‑to‑end, a job that previously stalled midway.
Safety improvements accompany the performance boost. Anthropic’s blog notes a lower incidence of “undesirable behaviors” such as cooperation with misuse, deceptive replies, and hallucinations. The model more reliably refuses malicious requests and resists prompt‑injection attacks, though the company concedes it does not yet match the safety profile of Opus 4.8 or Claude Mythos Preview on high‑risk tasks.
Anthropic will make Sonnet 5 the default model for all free and Pro subscription plans, signaling the company’s confidence that agentic capability is now a standard expectation across price tiers. The launch follows similar moves by OpenAI, which previewed GPT‑5.6 Sol, and Google, which rolled out Gemini 3.5 Flash as an agentic tool for planning and iteration.
Industry observers see the rollout as a clear marker that the competition is shifting from “who can do agentic work best” to “who can do it cheapest and most reliably.” With Sonnet 5, Anthropic aims to give developers a lower‑priced option that still delivers high‑quality results, letting users balance cost against performance across a broader spectrum of AI‑driven applications.
Dieser Artikel wurde mit Unterstützung von KI verfasst.
News Factory APP - agentische News für besseres SEO & AEO.