Anthropic launches Claude Sonnet 5: near-Opus performance at a lower price

Anthropic launched Claude Sonnet 5, its mid-tier model, on June 30, 2026. Sonnet 5 replaces the previous Sonnet 4.6 and lands close to Anthropic's top model Opus 4.8 in performance while costing far less. Its introductory pricing, in effect through August 31, 2026, is $2 per million input tokens and $10 per million output tokens. ASAP summarizes it from Anthropic's official announcement and primary reporting by AWS and TechCrunch.

A new mid-tier model that pushes Sonnet 4.6 aside

What Anthropic shipped is Claude Sonnet 5, a new mid-tier model in the Sonnet line that clearly takes over from the previous flagship, Sonnet 4.6. Anthropic says Sonnet 5 shows substantial improvement over its predecessor on core agentic capabilities such as reasoning, tool use, coding, and knowledge work. That list is not random; it names exactly the places where agents actually earn their keep today. Rather than leading with a headline benchmark score, Anthropic foregrounds the practical work of calling tools and fixing code. On Amazon Bedrock it is callable from launch day under the model ID us.anthropic.claude-sonnet-5.

A two-stage price: $2 intro, $3 standard

Sonnet 5 carries introductory pricing of $2 per million input tokens and $10 per million output tokens through August 31, 2026. After the introductory period it moves to standard pricing of $3 per million input tokens and $15 per million output tokens. VentureBeat framed this as a steep discount relative to Anthropic's top model, an aggressive pricing move made as the company races toward a blockbuster IPO.

Near Opus 4.8, and lower error rates

Sonnet 5's performance is close to that of the top model Opus 4.8 while staying at a lower price, according to Anthropic. The company says Sonnet 5 lowers hallucination and sycophancy rates and improves safety metrics measured in behavioral audits. Specific benchmark figures are detailed separately in the Claude Sonnet 5 System Card, while the announcement presents comparison numbers only as a table image.

Why the price is the real story

Sonnet 5's launch shows Anthropic extending the performance race into a price race. Pulling Opus 4.8-class intelligence down to Sonnet-tier pricing directly cuts token costs for enterprise customers that run agents at scale in repeated loops. As automation workloads that were too expensive on the top model shift down to a mid-tier model, the deciding factor moves from absolute performance to cost-performance. This is where the lower error rates matter. In high-volume loops a single bad response spreads into retries and review work, so cheaper tokens only lower the effective unit cost when they are paired with fewer mistakes.

What practitioners should check right now

The first thing to watch is the clock on that introductory price. The $2 rate holds only through August 31, 2026, after which it rises to $3, so any cost you measure during this window will drift if you drop it straight into an annual budget. Run your benchmarks during the intro period, but size your budget against the standard $3 and $15 rates. The word "close" also deserves scrutiny rather than trust. Because the announcement gives numbers only as a table image and defers the details to the System Card, the real gap versus Opus 4.8 is something each team has to remeasure on its own tasks. Drawing the line between workloads that can safely move down to Sonnet 5 and those that should stay on Opus is the core exercise this launch sets up.

Source: Anthropic official announcement (2026-06-30, "Introducing Claude Sonnet 5"), reporting by AWS, TechCrunch, and VentureBeat.