Anthropic SDK for Python v1.12.0 adds support for Claude Haiku 5.5 and new model capabilities

Anthropic released version 1.12.0 of its official Python SDK, adding support for the new Claude Haiku 5.5 model and introducing typed toolsets for computer and browser interactions. These toolsets are designed to streamline agentic workflows by providing structured interfaces for tool use, which can improve reliability and developer experience when integrating model outputs with external systems.

The update also adds web search and code execution capabilities to the list of supported model features, expanding the range of tasks that developers can orchestrate through the SDK. Additionally, the release introduces new fields for RBAC roles, including a `display_name` attribute and deprecation of the `name` field, which may reflect internal changes to role management in Anthropic’s APIs.

New lifecycle stage fields and a filter for the `/v1/models` endpoint are included, allowing developers to programmatically select models based on their operational stage. A new `include_default` parameter for listing workspaces provides finer control over workspace queries, and a `line` field has been added to model objects to support additional metadata.

Bug fixes in this release address the sending of the beta header by default when listing spend limits and ensure empty strings are correctly serialized in query parameters and multipart forms. Another fix prevents loss of local edits after a memory upload when the response is interrupted.

Documentation updates clarify rate limit list descriptions, managed agents’ multiagent and thread behaviors, and the definition of the new `line` field in model objects. The changelog also notes internal refactoring and test suite improvements, including the removal of skips for previously failing tests due to query parameter bugs and incomplete example responses.