NEWVectors or files. Pick a path.Start →
    Not runnable todayEmbeddingVerified 2026-11-05

    MobileCLIP (S0, S1, S2, B)

    apple/MobileCLIP-S2

    MobileCLIP (S0, S1, S2, B) is a image and text embedding model (apple-amlr). Not runnable on Mixpeek today.

    Something blocks it today, such as the license, remote code or a GPU-only runtime. The reason is below.

    Record as JSON

    No demo for MobileCLIP (S0, S1, S2, B) yet

    A video, image, slideshow or playground goes here once we record one on our own data.

    What it is good and bad at

    Best at
    Very fast image-text embedding on CPU and phones (S0 is the smallest) at CLIP-level zero-shot accuracy.
    Worst at
    Cannot be used commercially; lower ceiling than SigLIP so400m.
    Task
    Embedding
    Input
    Image, Text
    Output
    Vector
    Use cases
    Visual search, Product search

    Templates that use it

    No template uses this model yet. Browse templates.

    Input and output

    Input

    image
    resolution
    256x256
    color
    RGB
    text
    max_tokens
    77
    prompts
    CLIP-style captions

    Output

    vector
    dims
    512
    distance
    cosine
    type
    dense

    How Mixpeek wires it

    Extractors
    None today
    Feature URIs
    None today
    Vector indexes
    None
    Retriever stages
    None
    Combines with
    • CPU frame embedding at scale to zero, if the license allowed it

    How it runs

    Status
    Not runnable today Blocked by license (research only). Technically it would run as a CPU plugin through open_clip (the -OpenCLIP repos ship open_clip_model.safetensors); the mobileclip repos ship .pt pickles.
    Serve apps
    None
    Runtime
    open_clip (MobileCLIP-*-OpenCLIP) or the mobileclip package
    Weights
    safetensors, no remote code
    Parameters
    Not recorded
    Size on disk
    398 MB
    Preprocessing
    Resize and center-crop to 256; no mean/std normalization in the reference transform.
    Postprocessing
    L2-normalize; cosine similarity.
    Serving
    both, batch 64
    Hardware
    Runs on CPU (slower without a GPU)

    Scale to zero

    Minimum replicas
    0
    Maximum replicas
    1
    Idle timeout
    600 s
    Memory
    1 GB
    CPUs
    1
    GPUs
    0
    Cold start, node warm
    Not recorded
    Cold start, new node
    Not recorded

    Measured cost

    Unmeasured

    No committed measurement; plan row 3 measures it in the standalone image.

    Our evaluations

    Unmeasured

    We have not evaluated this model on our own data yet. Publisher results are in the sources below.

    Questions

    What is MobileCLIP (S0, S1, S2, B) best at?
    Very fast image-text embedding on CPU and phones (S0 is the smallest) at CLIP-level zero-shot accuracy.
    What are the limits of MobileCLIP (S0, S1, S2, B)?
    Cannot be used commercially; lower ceiling than SigLIP so400m.
    Can MobileCLIP (S0, S1, S2, B) be used commercially?
    No. Apple Machine Learning Research Model license: use only for Research Purposes, which exclude commercial exploitation, product development or use in any commercial product or service.
    How does MobileCLIP (S0, S1, S2, B) run on Mixpeek?
    Blocked by license (research only). Technically it would run as a CPU plugin through open_clip (the -OpenCLIP repos ship open_clip_model.safetensors); the mobileclip repos ship .pt pickles.
    What does MobileCLIP (S0, S1, S2, B) output?
    A 512-dimensional dense vector, compared by cosine similarity.