Claude Sonnet 5 becomes the default across all Claude plans
By Steven Van ·
Anthropic's new agentic Sonnet nears Opus 4.8 performance at a lower price and is now the default for Free and Pro users.
On June 30, 2026, Anthropic introduced Claude Sonnet 5, which it calls the most agentic Sonnet model yet. The pitch is straightforward: a model that makes plans, drives tools like browsers and terminals, and runs autonomously at a level that until recently demanded larger, more expensive models. It is a notable release because Sonnet 5 is the source model that many downstream tools quickly adopted after launch.
Built to run agents, not just answer prompts
Where earlier Sonnet models were framed around fast, capable responses, Sonnet 5 is designed around autonomy. Anthropic positions it as a model that can break a goal into steps, call tools such as web browsers and command-line terminals, and carry a task forward on its own without constant hand-holding. That shift matters for anyone building agents: multi-step workflows that once needed a top-tier flagship to stay coherent can now lean on a mid-tier Sonnet, which changes the economics of running agents at scale.
Near-flagship quality at a lower price
The headline trade-off is cost. According to Anthropic and outlets covering the launch, Sonnet 5's performance lands close to Anthropic's larger Opus 4.8 model while costing considerably less to run. Anthropic reported an agentic-coding benchmark score of 63.2% for Sonnet 5, sitting between Sonnet 4.6 at 58.1% and Opus 4.8 at 69.2%. On pricing, Sonnet 5 launched with introductory rates of $2 per million input tokens and $10 per million output tokens through August 31, 2026, moving afterward to $3 and $15 per million tokens. The framing from press coverage was consistent: a cheaper way to run agents without dropping to a noticeably weaker model.
The new default across Claude plans
Availability is broad. Anthropic made Sonnet 5 available across all plans and set it as the default model for Free and Pro users, with access extended to Max, Team, and Enterprise customers as well. Making a new agentic model the default for free and paid consumer tiers is a meaningful signal — it means most everyday Claude interactions now run on a model tuned for planning and tool use rather than one tuned purely for chat.
Safety framed as an agentic concern
Because the model is built to act autonomously, Anthropic paired the release with a system card and safety assessment. The company said Sonnet 5 shows an overall lower rate of undesirable behaviors than Sonnet 4.6 and is generally safer to use in agentic contexts. That emphasis reflects the reality of agentic models: when a system can browse, run commands, and chain actions together, safety is less about a single reply and more about how the model behaves over a long, unsupervised sequence of steps.
Sources: Anthropic: Introducing Claude Sonnet 5, TechCrunch, Claude Sonnet 5 System Card.