Breaking

Anthropic Releases Claude Sonnet 5, Pitching Opus-Level Agentic Skills at Lower Cost

June 30, 2026 at 14:38 EDT

Foundation Models
AI Agents
Software Dev & Coding

Anthropic on June 30, 2026 unveiled Claude Sonnet 5, the latest in its Sonnet line, calling it its most agentic Sonnet yet and saying it can plan, use tools and carry out autonomous multi-step work at a level rivaling larger, more expensive models.

June 30, 2026 · Anthropic

Claude Sonnet 5 Ships: Opus-Class Agents at Sonnet Prices

Anthropic's "most agentic Sonnet yet" launches immediately — strengthened planning, tool use, and autonomous long-horizon execution, now the default model across every Claude plan.

token context window — long docs & large codebases

90%

cost cut via prompt caching

Day 0

available immediately + System Card published

Output price per 1M tokens

Same agentic ambitions as the flagship — at a fraction of the cost.

$10

Sonnet 5

intro · until Aug 31

$15

Sonnet 5

standard

$25

Opus 4.8

flagship

Input pricing follows the same gap: $2 / $3 (Sonnet 5) vs $5 (Opus 4.8) per 1M tokens.

Built to run autonomous agent loops

Plan

→

Use tools
browser · terminal

→

Execute
multi-step

→

Self-check

Native high-res vision, a new efficient tokenizer, fewer unnecessary refusals, and stronger resistance to prompt injection in agentic scenarios.

Developer appeal

Approaches Opus 4.8 on planning and long-horizon tasks — at Sonnet pricing. Higher completion rates reported in coding, debugging, and browsing, with the model finishing tasks that previously stalled.

Caveats

Large-scale long-term evaluation is still pending so soon after launch. Some still recommend Opus or Mythos for extremely complex tasks, and safety claims await real-world verification.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.