Anthropic has released Claude Sonnet 5, calling it the most agentic Sonnet-class model the company has built. Unveiled on June 30, 2026, the new model can make plans, use tools like browsers and terminals, and run autonomously at a level that, just a few months ago, required larger and far more expensive models. The launch comes at a moment when the entire AI industry is racing to turn chatbots into capable agents that can act on a user's behalf rather than merely answer questions.
For developers tracking the latest developments, the news adds to a busy stretch of breaking AI news that has reshaped the competitive landscape. Anthropic frames Sonnet 5 as the moment agentic capability finally arrives at mid-tier prices, a shift that could change how teams budget for autonomous workflows.
Performance That Rivals the Flagship
According to Anthropic, Sonnet 5 narrows the gap between the mid-range Sonnet line and the company's top-tier Opus models. Its performance is close to that of Opus 4.8, but at substantially lower prices, and it represents a marked improvement over its predecessor, Sonnet 4.6, on reasoning, tool use, coding, and knowledge work.
Anthropic published cost-performance curves comparing the model against Sonnet 4.6 and Opus 4.8 on two agentic evaluations: BrowseComp, which measures agentic search, and OSWorld-Verified, which tests computer use. The company says Sonnet 5 offers a wider range of cost-performance options than Sonnet 4.6 and, at higher effort levels, can match Opus 4.8 on some tasks.
Aggressive Pricing for Agents
Pricing is central to the pitch. Claude Sonnet 5 launches with introductory pricing of $2 per million input tokens and $10 per million output tokens, available through August 31, 2026. After that window, it moves to $3 per million input tokens and $15 per million output tokens.
For comparison, Opus 4.8 is priced at $5 per million input tokens and $25 per million output tokens. Because agents burn through large numbers of tokens as they reason and call tools over many steps, that price gap matters enormously for enterprise budgets. Early-access testers cited by Anthropic described Sonnet 5 finishing complex tasks where previous Sonnet models would stop short, and checking its own output without being explicitly asked.
Availability Across Every Plan
Claude Sonnet 5 is available across all of Anthropic's plans from launch day. It is the default model for Free and Pro users, and it is available to Max, Team, and Enterprise customers. Developers can access it through the Claude API using the model identifier `claude-sonnet-5`, as well as through Claude Code and the Claude Platform.
The broad rollout is notable because it puts capable agentic performance into the hands of free users, not just paying enterprise customers. Anthropic said the Sonnet line has historically been where the agentic era began for many developers, pointing to Claude Sonnet 3.5, 3.6, and 3.7 as the first models that showed strong coding and tool-use skills.
A Safety Story Built In
Safety was a key part of the announcement. Anthropic said its 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. The company also noted that the model has a much lower ability to perform cybersecurity tasks than its current Opus models, a deliberate design choice that reduces the risk of misuse even as general capabilities climb.
The Claude Sonnet 5 System Card, released alongside the model, reports a broader set of evaluations in detail. Anthropic has positioned transparency around safety testing as a competitive differentiator as regulators in the United States and Europe scrutinize frontier model releases.
The Broader Agentic Moment
The launch lands amid intense competition. Rivals including OpenAI, Google, and a fast-moving set of Chinese labs are all pushing toward models that can take multi-step actions rather than simply generate text. By delivering Opus-adjacent agentic capability at Sonnet prices, Anthropic is betting that the next battleground is not raw benchmark scores but the cost of running an agent reliably over long, complex tasks.
With introductory pricing locked in through the end of August, developers now have a window to stress-test Sonnet 5 against their most demanding agentic workloads before the higher rates take effect.
Stay Ahead of AI
The pace of model releases shows no sign of slowing. For ongoing coverage of new launches, funding rounds, and policy shifts, visit AI Buzz Wire for the latest AI developments as they happen.
Read more AI news →
