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

The launch

Anthropic launched Claude Sonnet 5 on June 30, 2026. The company describes the new model, in its official announcement, as the most agentic Sonnet it has ever built, surpassing the previous version, Sonnet 4.6, in reasoning, tool use, coding and knowledge work. According to TechCrunch, Sonnet 5 became the default model on the Free and Pro plans of Claude.ai as soon as it launched, and was also immediately available to Max, Team and Enterprise subscribers, as well as in Claude Code and the Claude Platform.

A 1 million token context window and promotional pricing

The model ships with a native 1 million token context window, letting it process long documents and entire codebases in a single conversation without splitting the material into parts. The announced promotional pricing is $2 per million input tokens and $10 per million output tokens, valid through August 31, 2026, after which it rises to $3 and $15 respectively. According to Search Engine Journal, Anthropic positions Sonnet 5 as offering near-Opus intelligence, but accessible across every plan, including the free tier, reinforcing the company's strategy of popularizing its frontier models rather than reserving them for the highest-paying customers.

Agentic capability takes center stage

The technical highlight of the launch is the model's ability to plan multi-step tasks, use tools like a browser and a terminal, and operate autonomously in workflows that, just a few months ago, only ran reliably on larger, more expensive models, according to Anthropic itself. Independent evaluations are already circulating: developer Theo Browne tested the model on his own coding benchmarks and reported, according to coverage by BigGo Finance, a 37% success rate on a set of hard tasks, at a cost of roughly $6,000 in API calls, a reminder of how expensive it still is to run AI agents end to end on complex tasks, even with a cheaper model.

Why it matters for automation teams

For SMBs and teams automating customer service, marketing and operations with AI agents, a cheaper entry-level agentic model changes the math on which workflows are worth shipping. Teams that today reserve Opus only for critical tasks, because of cost, gain room to test more ambitious automations using Sonnet 5 as a middle tier, without giving up reasoning quality. It is Anthropic's most concrete bet yet on making frontier agentic AI accessible at scale, not just for those paying for the priciest plans. The underlying question remains: with Sonnet, Opus, Fable 5 and Mythos 5 all competing for space in the same portfolio, Anthropic looks less focused on selling one perfect model and more focused on having the right option for every budget.