Claude Sonnet 4.7: more reliable agents and longer task duration
In one sentence Anthropic updates Sonnet to 4.7: focused on agent reliability over long tasks, better tool use, tighter integration with Claude Code and the Claude Agent SDK.
After Sonnet 4.5 (September 2025) and Haiku 4.5 (October 2025), Anthropic releases Sonnet 4.7. It's not a generation jump — Opus 5 / Sonnet 5 are expected later in 2026 — but a targeted update to a line that has become the de facto default for agent builders.
The stated focus is "reliability on long tasks": fewer mid-session errors, less drift, better recovery when a tool fails. The kind of improvement that barely shows in demos but matters enormously in production workflows lasting hours.
Coding also improves: SWE-bench remains the ground where Anthropic and OpenAI trade leadership every few months. Sonnet 4.7 tries to take it back.
Pricing unchanged from 4.5. Rollout to Claude Code, GitHub Copilot, Cursor, Windsurf within hours — by now a sector ritual.
Companies
Anthropic
Tools
Claude Sonnet 4.7, Claude Code, Claude Agent SDK
Tags
Sources