Anthropic releases Claude Sonnet 5, its most agentic Sonnet model yet

The most agentic Sonnet yet

Anthropic introduced Claude Sonnet 5 on June 30, presenting it as the most agentic Sonnet model the company has built so far. It can make plans, use tools such as browsers and terminals, and run autonomously at a level that, only a few months ago, required larger and more expensive models.

According to Anthropic, the agentic AI era began with Sonnet-class models: Claude Sonnet 3.5, 3.6 and 3.7 were the first models that showed impressive skills in coding and tool use. More recently, the clearest gains in agentic capabilities came from the Opus-class models. Sonnet 5 narrows that gap: its performance is close to Opus 4.8, but at lower prices.

Benchmarks, effort levels and pricing

The company says Sonnet 5 is a substantial improvement over its predecessor, Sonnet 4.6, on important aspects of agentic performance such as reasoning, tool use, coding and knowledge work. It provides substantially improved cost efficiency at medium effort, while its higher-effort performance can match Opus 4.8 on some tasks. Users can adjust the effort level to find the right balance between cost and performance.

Claude Sonnet 5 is available across all plans from day one: it is the default model for Free and Pro plans and is available to Max, Team and Enterprise users. It is priced at $2 per million input tokens and $10 per million output tokens. That introductory pricing has since been made permanent, against standard pricing of $3 and $15. Anthropic notes that the model uses an updated tokenizer, so the same input can map to roughly 1.0 to 1.35 times as many tokens depending on the content type.

Safety evaluations

Anthropic's safety assessments found that Sonnet 5 shows an overall lower rate of undesirable behaviours than Sonnet 4.6 and is generally safer to use in agentic contexts. Evaluations also show that it has a much lower ability to perform cybersecurity tasks than the company's current Opus models. In a test that measures whether a model can develop exploits for vulnerabilities in the Firefox browser, Sonnet 5 was never able to develop a full working exploit, though it showed a slightly higher rate of partial success than Sonnet 4.6. Full results are published in the Claude Sonnet 5 System Card.

Developers quoted by Anthropic praised the model. Zimu Li said Sonnet 5 gives agents a strong execution layer for multi-step software engineering work, handling sustained coding, tool use and debugging well across messy technical contexts. Deepak Singh, VP at Kiro, described it as a strong agentic-coding model with top-tier accuracy comparable to Opus-class models and a clear step-function improvement over Sonnet 4.6.