More than a checkpoint update
Anthropic introduced Claude Sonnet 4.5 on September 29, 2025. The announcement included the Claude Agent SDK and product changes such as checkpoints and tools for managing longer-running work.
The industry significance was the growing importance of the runtime around a model: how it receives context, uses tools, stores progress, and recovers from mistakes.
Recovery is a feature users notice
An agent that can complete a task only when nothing interrupts it is difficult to rely on. Users need to inspect progress, stop a run, and resume without losing accepted work or repeating a costly action.
For a creative operation, this can mean preserving an approved brief while regenerating only the rejected asset. For software work, it can mean restoring a known code state after an unsuccessful experiment.
Evaluate the complete system
Test a representative task with a deliberate interruption and a tool failure. Check what state survives, what the user can review, and how the system avoids duplicate work. Compare the final artifact with the original requirements. A stronger model helps, but dependable execution also comes from the surrounding controls and the clarity of the task it has been given.
Official sources
This article covers an AI industry event. XMH.NET specializes in image generation and editing APIs; coverage does not imply that every model, product, or feature described is available through our service.