Particle.news
Download on the App Store
23 ARTICLES
·
1h ago

Anthropic Launches Claude Sonnet 4.5, Claiming Coding Lead and 30-Hour Autonomous Runs

Anthropic targets enterprise adoption by pairing benchmark-topping coding with stricter safety.

Abstract composition with flowing head silhouette and stylized hand gesture, holding geometric cube structure
Anthropic logo is seen in this illustration created on May 20, 2024. REUTERS/Dado Ruvic/Illustration/File Photo
Image

Overview

  • Claude Sonnet 4.5 is now available in the Claude apps and via API at the same Sonnet pricing, and Anthropic is making it the default model for most use cases.
  • The model posts a 77.2% score on SWE-Bench Verified (82% with parallel test-time compute) and 61.4% on OSWorld, results Anthropic says surpass its Opus 4.1 and some competitor systems.
  • Internal and customer trials reported autonomous operation for about 30 hours on complex, multistep tasks, up from roughly seven hours for Opus 4.
  • Anthropic says Sonnet 4.5 is its most aligned release to date, shipped under AI Safety Level 3 with reduced sycophancy, deception and power-seeking plus stronger prompt-injection defenses.
  • The launch adds Claude Code checkpoints, a refreshed terminal and a native VS Code extension, a Claude Agent SDK, and new memory and context tools, while Microsoft announced new 365 Copilot features powered by Anthropic models.