Python 3.15 RC1: Lazy Imports, frozendict, and Tachyon Profiler

Python 3.15 Arrives at Release Candidate Stage

On 4 August 2026, the Python core development team released Python 3.15.0rc1, the first release candidate for what will become the October 2026 stable release. A release candidate means the feature set is now frozen — no new functionality will be added before the stable release. The two remaining milestones are a second release candidate (3.15.0rc2, scheduled for 1 September 2026) for final bug validation, then the stable 3.15.0 on 1 October 2026. Python 3.15 carries the most significant startup performance improvement in the language's history for large codebases, a new built-in immutable mapping type, a bundled high-frequency profiler, JIT compiler improvements of 8 to 13 per cent across architectures, and a platform-universal default encoding change to UTF-8 under PEP 686. Teams using Python for backend services, data pipelines, and AI application development should begin testing their codebases against the release candidate now to surface any incompatibilities before the stable October release.

Lazy Imports: Up to 70 Per Cent Faster Startup for Large Codebases

The headline feature of Python 3.15 is lazy imports, specified under PEP 810, which adds a new lazy keyword for deferred module loading. When a module is imported using the lazy keyword, Python defers loading it until the first time a name from that module is actually accessed at runtime. For applications that import many modules at startup — the common pattern in frameworks, Django and FastAPI services, data science toolkits, and large CLI tools — lazy imports can dramatically reduce the time Python spends importing before the application is ready to serve requests. Meta, which contributed to the PEP development process, reported startup time reductions of up to 70 per cent on large Python codebases when lazy imports are applied to the heaviest import chains. For teams running Python services in autoscaling environments where container cold-start time directly affects the response time floor — including Kubernetes deployments and AWS Lambda Python runtimes — this is a structural improvement rather than an incremental one. Python 3.14 offered no comparable mechanism for deferred module loading at the language level, requiring teams to implement manual lazy loading patterns or third-party deferred-import packages as workarounds.

Built-in frozendict: A Native Immutable Mapping Type

Python 3.15 introduces a built-in frozendict type — an immutable dictionary that, like frozenset, is hashable and can be used as a dictionary key or a set member. Prior to 3.15, developers needing an immutable mapping had to use third-party packages from PyPI, convert data to a tuple of key-value pairs, or use types.MappingProxyType from the standard library — none of which provided the clean semantic equivalence of a hashable dict literal. The built-in frozendict covers the most common production use cases: configuration objects that should not be mutated after construction, cache keys derived from mapping contents, and functional programming patterns that require immutable data structures. For teams building strongly-typed, defensively written Python APIs — increasingly the norm in Indian software teams shipping production-grade backends — frozendict removes a consistent source of improvised workarounds.

The Tachyon Profiler: Continuous Performance Visibility

Python 3.15 bundles the Tachyon Profiler as a new standard-library package. Tachyon uses a high-frequency statistical sampling approach that significantly reduces profiling overhead compared to the existing cProfile and profile modules in the standard library. The lower overhead makes it more practical to run Tachyon continuously in development and staging environments rather than as a targeted diagnostic step, surfacing performance regressions as they are introduced rather than after they accumulate into a production incident. Teams that have historically avoided continuous profiling because of the overhead cProfile introduces into test and development runs will find Tachyon a meaningful addition to their standard toolbox.

JIT Improvements and the UTF-8 Default Encoding Change

Python 3.15 delivers JIT compiler speedups of 8 to 13 per cent depending on CPU architecture, continuing the multi-release improvement arc that began when the experimental JIT compiler was introduced in Python 3.13. The range reflects differences in how the JIT's generated machine code interacts with each architecture's cache hierarchy and branch predictor — ARM-based hardware, including Apple Silicon and AWS Graviton instances, generally benefits at the upper end of this range for workloads that stress tight loops and numeric operations. Under PEP 686, Python 3.15 changes the default text encoding to UTF-8 across all platforms. On Windows, previous Python versions defaulted to the system locale encoding — often CP1252 or a regional code page — which caused silent encoding mismatches when code or files were moved between platforms or consumed content from different systems. UTF-8 as the universal default eliminates an entire class of hard-to-diagnose encoding bugs that are particularly common in codebases maintained across Indian and European developer teams where machine locale settings differ.

What Python 3.15 Means for Indian Development Teams

For Indian Python development teams — where the language underpins backend services, data engineering pipelines, machine learning model serving, and large Django and FastAPI applications — Python 3.15 delivers improvements across two practically important dimensions. Startup latency and the UTF-8 encoding change are the highest-priority areas to test before the stable October release. Teams running large Python services in AWS Graviton or Kubernetes autoscaling environments should measure the startup time improvement from lazy imports against their specific import patterns — codebases with twenty to fifty top-level imports at startup are the most likely to see the largest absolute reductions. The frozendict and Tachyon additions are lower urgency but meaningful quality-of-life improvements for teams investing in clean, defensively written Python codebases. Teams currently on Python 3.12 or 3.13 should plan their upgrade path now given the October 2026 stable release timeline.

The Bottom Line

On 4 August 2026, the Python core development team released Python 3.15.0rc1, the first release candidate for the October 2026 stable release. The feature set is frozen, with RC2 scheduled for 1 September and the stable release for 1 October 2026. Python 3.15's headline features are lazy imports under PEP 810 — delivering up to 70 per cent startup time improvements on large codebases — a built-in frozendict immutable mapping type, the Tachyon high-frequency profiler, JIT improvements of 8 to 13 per cent across architectures, and UTF-8 as the platform-universal default encoding under PEP 686. For Indian Python development teams, startup latency improvements and the UTF-8 encoding change are the two highest-priority areas to evaluate against existing codebases before the stable October release.