Pydantic v2.10 introduces several new features and bug fixes, with a focus on enhancing functionality before pivoting to performance improvements in the next release. Key updates include support for partial validation, which is beneficial for processing incomplete data streams, and default factories that take validated data as arguments. The release also supports typing.Unpack, compiled patterns in protected namespaces, and the defer_build setting for TypeAdapter and Pydantic dataclasses, which improves startup performance. Additional enhancements include support for fractions.Fraction, improved JSON schema generation, and compatibility warnings for mixing v1 and v2 models. While performance was not the main focus, some optimizations were made, such as better namespace management and revalidation of parametrized generics. Notable changes include relaxing the protected_namespace config default to accommodate common field names in AI and data science, deprecating the schema_generator config argument for future performance improvements, and adopting modern functions for Base64Bytes and Base64Str types. The release also tidies up JSON schema generation for Literals and Enums and extends date support to 1BC. Pydantic v2.10 is highlighted as the most feature-rich version yet, thanks to contributions from over 30 contributors, and the team invites feedback and bug reports via GitHub discussions and issues.