AlphaCode 2 is for contest coding; AlphaEvolve is for algorithm discovery
Google DeepMind’s AlphaCode 2 and AlphaEvolve are aimed at different coding problems. The AlphaCode 2 technical report, dated 2023-12-06, describes a competitive-programming system built on Gemini models plus bespoke search, filtering, clustering, and reranking; DeepMind says it was evaluated on the same platform as the original AlphaCode and reached the 85th percentile on average there. AlphaEvolve, announced on 2025-05-14, is a Gemini-powered coding agent for general-purpose algorithm discovery and optimization. DeepMind says it pairs Gemini models with automated evaluators and an evolutionary framework, and that it has been used on math and computing problems such as matrix multiplication and data-center, chip-design, and AI-training efficiency work. This briefing stays within those documented scopes; it does not establish a single overall winner.
Check the sources
This is a NextGen-AI briefing, not a reproduction of the discovery source. A newsletter may cover several developments; consult the evidence below for this briefing’s specific subject. The source date refers to the discovery entry, not necessarily the announcement date.