Full story
Claude Sonnet 5 Launch
Anthropic introduced Claude Sonnet 5 as a more affordable model that narrows the gap between Sonnet and Opus, calling it "its most agentic Sonnet model to date."
“its most agentic Sonnet model to date”
The company says Claude Sonnet 5 can make plans, use tools like browsers and terminals, and run autonomously, while Opus models have better agentic capabilities but are more expensive than Sonnet models.
MacRumors reported that Sonnet 5’s performance is similar to Opus 4.8 and that it improved over Sonnet 4.6 in reasoning, tool use, coding, and knowledge work.
In Anthropic’s own announcement, the company said its safety assessments found that Sonnet 5 shows "an overall lower rate of undesirable behaviors than Sonnet 4.6" and is generally safer to use in agentic contexts.
Anthropic also set the rollout so that from today Claude Sonnet 5 is available across all plans, including being the default model for Free and Pro plans.
Pricing and Availability
Anthropic set introductory API pricing for Claude Sonnet 5 at $2 per million input tokens and $10 per million output tokens through August 31, after which prices will go up to $3 and $15 respectively.
Thurrott said the introductory pricing would last through August 31, 2026, when pricing shifts to $3 per million input tokens and $15 per million output tokens.
The GitHub Blog said Claude Sonnet 5 is now available in GitHub Copilot and can be selected in the model picker in Visual Studio Code, Visual Studio, Copilot CLI, and GitHub Copilot cloud agent.
In Anthropic’s announcement, the company said Claude Sonnet 5 is available to Max, Team, and Enterprise users and is available in Claude Code and on the Claude Platform.
VentureBeat reported that Sonnet 5 is the default model for users on Anthropic's Free and Pro plans while also being available to Max, Team, and Enterprise customers.
Agent Performance and Limits
Anthropic said testers described Claude Sonnet 5 finishing complex tasks "where previous Sonnet models would stop short," and also described it checking its own output without explicitly being asked.
“where previous Sonnet models would stop short”
In the same announcement, Anthropic said it handed Claude Sonnet 5 a two-part job—"update Salesforce account tiers, send a launch announcement to enterprise contacts"—and that it finished end to end, while "That used to stall halfway."
VentureBeat reported benchmark results including that on SWE-bench Pro, Sonnet 5 scores 63.2% compared with Sonnet 4.6’s 58.1% and Opus 4.8’s 69.2%.
The New Stack said Anthropic stresses that Opus 4.8 will still offer higher accuracy, and it noted that Sonnet 5 provides developers with lower-priced options that are "of much higher quality than what was previously available."
Anthropic also stated that it has a much lower ability to perform cybersecurity tasks than its Opus models, and it said Sonnet 5 "didn’t deliberately train Sonnet 5 on cybersecurity tasks."
