openai/CLIP
Read the upstream summary on the left, browse the cached forks below it, and load each fork comparison into the right-hand panel.
openai/CLIP
openai/CLIP is a Python repository for Contrastive Language-Image Pretraining: a model that takes image/text pairs and predicts the most relevant text for an image. It is active, not archived, and still receiving updates as of 2026-03-25. The repo is fairly popular upstream, with 33,015 stars and 3,972 forks.
Jump straight into Discofork's strongest cached fork picks, or open a compare view in one click.
Choose a fork to inspect
Choose this fork if you want a maintained CLIP codebase with extra prompt/tokenizer workflow support and stronger repo hygiene. Stick with upstream if you want the simplest, most canonical CLIP package and do not need the fork-specific prompt or tokenizer changes.
Choose this fork if you care most about packaging, publishing, and compatibility maintenance. Choose upstream if you want the latest CLIP codebase and ongoing updates.
Prefer upstream unless you specifically need this fork’s repository state for local experimentation or pinning. It does not show meaningful added capabilities, and it is slightly behind upstream, so it is best treated as a maintenance copy rather than a differentiated distribution.
Prefer upstream unless you specifically need this older snapshot; this fork adds no visible capabilities and lags upstream maintenance.
Choose upstream unless you specifically want a dormant personal fork with no functional changes. This fork adds no clear capabilities and is behind on maintenance, so it is not a better default for active use.
Prefer this fork only if you specifically want its performance-oriented, CPU/CUDA artifact-heavy workflow. For most adopters, upstream is the safer choice because this fork is much older and likely missing many later fixes and maintenance updates.
Choose this fork if your priority is MindSpore adoption and you want conversion/documentation help; choose upstream if you want the maintained PyTorch implementation and broader compatibility.
Prefer upstream unless you specifically want this exact frozen snapshot. The fork adds no visible features and is behind upstream, so it offers little advantage for new adoption.
Prefer this fork only if your main goal is ImageNet evaluation and you want the added script and labels in one place; otherwise upstream is the better default because this fork is old and substantially behind.