technicolor-research/faiss-quickeradc
stale
significant_divergence
Selected Choose this fork if your main goal is QuickerADC-style product-quantization speedups and you can live with a stale, heavily diverged codebase. Choose upstream FAISS if you want current maintenance, broader feature coverage, and lower integration risk.
efficient/faiss-learned-termination
stale
significant_divergence
Prefer this fork if you want the learned adaptive early termination research stack and its reproducibility assets. Prefer upstream if you need an actively maintained, current Faiss for production or integration work.
bitsun/faiss
stale
significant_divergence
Prefer this fork only if Windows/cross-platform portability is the primary requirement and you can absorb maintenance debt. For most adopters wanting current Faiss behavior, upstream is the better default because this fork is significantly behind and appears stale.
j40903272/faiss
stale
significant_divergence
Prefer this fork only if you need an older, self-contained Faiss snapshot with generated wrappers/docs and are willing to give up current upstream fixes and API evolution. For active development or production use, upstream is the better choice.
VinhASOKEN/faiss
stale
significant_divergence
Choose this fork only if you need its specific local changes and can afford to own long-term merge and validation work. If you want current Faiss behavior, performance fixes, and API compatibility, upstream is the safer default.
zhou-yuxin/optimized-faiss
stale
significant_divergence
Choose this fork only if the extra CPU/SIMD optimization work is the main requirement and you can absorb ongoing maintenance. If you want current Faiss features, validation fixes, and a lower-risk upgrade path, upstream is the better default.
blevesearch/faiss
active
significant_divergence
Choose this fork if your priority is the extra IVF/binary-index functionality and merge-oriented workflows; stick to upstream if you want the latest Faiss fixes, broader compatibility, and less maintenance risk.
mdouze/faiss
stale
significant_divergence
Prefer this fork only if you need its specific IO/API and compatibility changes. For most adopters, upstream is the safer default because this fork is materially behind and highly divergent.
ahojnnes/faiss
slowing
significant_divergence
Prefer this fork if you need its custom SIMD/quantization/validation work and can absorb long-term merge debt. Prefer upstream if you want the current Faiss API, active maintenance, and the broadest supported workflow surface.
infiniflow/faiss
stale
significant_divergence
Choose this fork only if its RAFT/CAGRA and benchmark-oriented additions are specifically valuable. If you want the safest, most current Faiss for production or long-term maintenance, upstream is the better default.