Anthropic Releases Claude Sonnet 5 with Enhanced Agentic Capabilities

Anthropic launched Claude Sonnet 5 on June 30, 2026, replacing the four-month-old Sonnet 4.6 as its medium-sized model. The company is offering introductory pricing of $2 per million input tokens and $10 per million output tokens through August 31, after which pricing increases to $3 and $15 respectively.

Sonnet 5 is now the default model for both free and Pro Claude users across the Chat, Cowork, Claude Code, and Platform products.

Performance Claims

According to Anthropic, Sonnet 5 delivers performance "close to that of Opus 4.8" while maintaining lower pricing. The company claims "substantial improvement" over Sonnet 4.6 in reasoning, tool use, coding, and knowledge work.

The 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 more expensive models," Anthropic states. Specific benchmark scores have not been disclosed.

Model Lineup Context

In Anthropic's three-tier system, Sonnet sits between the smaller Haiku model and the larger Opus model. The company also operates two other model lines: Mythos 5 (with reduced guardrails for security researchers and platform owners) and Fable 5 (the safer version currently blocked by U.S. government restrictions).

Claude Opus 4.8 was released in late May 2026. Sonnet 4.6, the previous medium-sized model, launched in February 2026.

Rate Limits and Availability

Anthropic increased rate limits across its products to accommodate higher token usage at elevated effort levels. Users can select effort levels based on project requirements. The model is available immediately through all Claude interfaces.

What This Means

Sonnet 5 represents Anthropic's attempt to compress flagship-level performance into a mid-tier price point. The four-month release cycle suggests aggressive iteration on agentic capabilities. The introductory pricing—50% below standard rates—indicates either competitive pressure or an effort to build market share before the price increase. Without published benchmarks, however, the claimed performance parity with Opus 4.8 remains unverified. The August 31 pricing cliff could create migration challenges for cost-sensitive applications.