TechCrunch: Anthropic launches Claude Sonnet 5 as a cheaper way to run agents

Anthropic released Claude Sonnet 5 on June 30, 2026, making it the default model for free and Pro plan users immediately. The model replaces Sonnet 4.6 as the workhorse tier, bringing agentic capabilities—autonomous planning, multi-step tool use, browser control—to a price point that was previously reserved for more expensive models.

For designers working with AI-assisted workflows, the practical implication is that Figma’s MCP integration, Claude Code roundtrips, and agentic design tasks now run on a significantly more capable model without a cost increase. Sonnet 5 achieves 63.2% on agentic coding benchmarks and performs on par with Opus 4.8 on knowledge work tasks, while coming in at $2 per million input tokens through August 2026.

The model’s improved tool use and planning make it better at multi-step tasks: iterating on a design system component, running a test loop across variants, or managing the back-and-forth between Figma context and code output. A Zapier engineer quoted in the TechCrunch piece noted that Sonnet 5 completes multi-step workflows “end to end” that previously stalled mid-task.

Anthropic also reports lower rates of deception, hallucination, and undesirable cooperation with Sonnet 5 compared to its predecessor—relevant for teams using agents in production-adjacent contexts like generating copy or managing design tokens at scale.