The Continuity Layer
Run your own AI · Fits Server

Open models that need a server

thecontinuitylayer.com/open/models/server

Checked 6 November 2026. US dollars. No affiliate links.

5 open models in our catalog are too big for one home machine and run on rented GPUs or a host: DeepSeek V4, GLM, Kimi, Muse, Mistral Large. 1 of them carry a permissive license (Apache 2.0 or MIT), 0 read images. Each figure comes from the maker's page or model card, checked 6 November 2026.

Server: rented gpus or a host: too big for one home machine. Want to filter by license, who set the behavior, or images? Use the model finder.

  • DeepSeek-V4.1-Flash

    Server-size

    DeepSeek · 552B parameters, 8–16B active

    Frontier-class open model; rent it through a host or router

    License
    MIT
    License rules
    None beyond keeping the notices
    Behavior
    Maker's own tuning
    Reads
    Text
    Context
    1M
    Download
    —

    Source · checked 3 November 2026 · DeepSeek V4 graded

  • GLM-5.3

    Server-size

    Z.ai · 753B parameters

    Coding agents through a cheap plan or a router

    License
    GLM-5.3 License (MIT-style)
    License rules
    Very large model-as-a-service operators must pass Z.ai's security review
    Behavior
    Maker's own tuning
    Reads
    Text
    Context
    Not stated on the card
    Download
    —

    Source · checked 3 November 2026 · GLM graded

  • Kimi K3

    Server-size

    Moonshot AI · 2.8T parameters, 104B active

    Long agent runs through a host or router

    License
    Kimi K3 License (MIT-style)
    License rules
    Model-as-a-service businesses over $20M a year need a separate agreement
    Behavior
    Maker's own tuning
    Reads
    Text
    Context
    1M
    Download
    —

    Source · checked 3 November 2026 · Kimi graded

  • Muse Spark Open

    Server-size

    Meta · 816B parameters, 42B active

    Meta's newest open release; needs serious hardware

    License
    Muse Model License
    License rules
    "Built with Muse Spark Open" display, Usage Policy, and a $20M model-as-a-service clause; gated download
    Behavior
    Maker's own tuning
    Reads
    Text
    Context
    1M
    Download
    —
    Note
    Meta says it is optimized to run on two Mac Studios with M5 Ultra.

    Source · checked 3 November 2026 · Muse graded

  • Mistral Large 4

    Server-size

    Mistral AI · About 1T parameters, 52B active

    Research and personal use at frontier scale

    License
    Mistral Ecosystem Open-Weight License
    License rules
    Commercial operation needs a separate license from Mistral
    Behavior
    Maker's own tuning
    Reads
    Text
    Context
    1M
    Download
    —

    Source · checked 3 November 2026 · Mistral Large graded

Other machine sizes

  • Fits 8GB · 7 models · most laptops, an 8gb mac, an 8gb graphics card
  • Fits 16GB · 11 models · a 16gb mac or laptop, a 12–16gb graphics card
  • Fits 24–32GB · 19 models · a 24gb graphics card, a 32gb mac
  • Fits 64–128GB · 22 models · a 64–128gb mac studio or 128gb mini pc

Not sure what your machine has? The will it run page says how to check, and the hardware guide says what to buy, or not.