Microbenchmark Performance Analysis: Python 3.15.0a3 vs 3.11.14

Microbenchmark Performance Analysis: Python 3.15.0a3 vs 3.11.14 Mehmet Keçeci ORCID: https://orcid.org/0000-0001-9937-9839 Received: 01.23.2026 Abstract: A comprehensive microbenchmark analysis comparing Python 3.15.0a3 (alpha) and Python 3.11.14 (stable) has been conducted on Linux/Conda-Forge platform. Eight runs (4 full + 4 summary benchmarks) per version measured 11 critical tests encompassing function calls, loop arithmetic, list/dict operations, string manipulation, mathematics, comprehensions, generators, recursion, LRU cache, attribute access, and exception handling. Key finding: Function calls dominate 99.9% of total execution time with no meaningful difference between versions (Py3.15: 61.934s, Py3.11: 61.820s, ratio=1.002). Microbenchmark-wise, Py3.11.14 demonstrates 64% superiority (7/11 tests). Comprehensions (1.338x) and generators (1.340x) exhibit 25-34% Py3.11 advantage, while Py3.15 excels in loops (0.821x) and attribute access (0.972x) by 2-18%. Linear scale visualization clarifies differences: Py3.11 leads in critical Python constructs (comprehensions, generators), while Py3.15 shows gains in fundamental arithmetic and attribute access. Standard deviation ranges 0.1-1.0%, indicating high measurement repeatability (95% CI ±1.5%). Practical implications: Python 3.11.14 recommended for production due to stability + microbenchmark superiority. Py3.15 alpha suitable for R&D projects where function calls are non-dominant. Comprehension/generator optimizations anticipated in Py3.15 final release. This study highlights Python evolution's function call dominance (99.9%) and microbenchmarks' disconnect from real-world performance. Keywords: Python microbenchmark, performance analysis, Python 3.15, Python 3.11, function calls, comprehension, generator, execution time, statistical analysis, benchmark repeatability.