Claude Sonnet 4.5 (AU)

InfoQ reported that Claude Sonnet 4.5 topped the SWE-Bench Verified benchmark, extending Anthropic's coding focus beyond 30 hours of sustained agentic work. The coverage highlights the model's positioning as a coding leader, building on the same launch details published by Anthropic on September 29, 2025. As third-party reporting, the InfoQ piece corroborates Anthropic's framing of Claude Sonnet 4.5 as a frontier coding model with practical endurance on complex, multi-step software tasks. It serves as independent confirmation of the SWE-Bench Verified result and the 30-hour sustained-task claim rather than introducing new technical claims.

Anthropic announced Claude Sonnet 4.5 on September 29, 2025, positioning it as its strongest model for coding, complex agents, and computer use. The launch shipped alongside upgrades including Claude Code checkpoints, a VS Code extension, context editing, a memory tool, and the Claude Agent SDK for developers. Pricing was set at $3/$15 per million tokens, unchanged from Claude Sonnet 4. On benchmarks, Claude Sonnet 4.5 reached state-of-the-art on SWE-bench Verified and was reported to maintain focus on multi-step tasks for more than 30 hours. On OSWorld, it led at 61.4%, up from Sonnet 4's 42.2% four months earlier. Anthropic also described it as the most aligned frontier model it had released, with improvements over prior Claude models.