Modelpublic

nanguoyu/DeepSeek-V4-Flash-0731-minirun

sourceHugging Facemitupdated 15d agoView on Hugging Face

0likes895downloads

Model Card

DeepSeek-V4-Flash-0731-minirun

The weights of `deepseek-ai/DeepSeek-V4-Flash-0731` at revision 7872f01b1d1fe23eabc4c98b48bffcef5a386062, repacked byte for byte into the container format Minirun reads. Minirun streams this model from an external SSD on Mac and iPhone through a memory budget you set, rather than loading it into memory, so the model does not have to fit in the machine. The reader is open source: `Sources/StorageCore/Container/` in `nanguoyu/minirun-app`.

What this is

A byte-preserving repack. No requantization, no retraining, no numerical change of any kind:

—the routed expert weights are the upstream FP4 packed values and their ue8m0 scales, copied verbatim and only reordered;
—the remaining quantized matrices are the upstream FP8 e4m3 values and their ue8m0 block scales, copied verbatim;
—all other tensors are the upstream BF16/F32/I64 bytes, copied verbatim.

Every byte of weight data in this repository is a byte of the source checkpoint at the pinned revision above, in a different order. The only bytes that are not are container headers, and zero padding where a container's scale region is shorter than its alignment unit.

The model configuration and tokenizer are verbatim files from the same pinned source revision. Their byte counts and SHA-256 identities are recorded in index.json.

Contents: 47 directories, 577 data files, 166.9 GB total, plus a per-directory manifest giving each file's shape, element width and offsets.

Run it with Minirun

Get the app. On a Mac, download `Minirun.dmg`. On an iPhone, install TestFlight and open testflight.apple.com/join/QuYTybjV on the phone; the source is at `nanguoyu/minirun-app` if you would rather build it.

Point it at this repository. In Settings → Storage, use Add a folder… to register a folder on an external NVMe drive. In Settings → Models, open Find Models, select this repository and press Download -- or point Minirun at a copy you already have. Run Verify all files, then start a chat.

What to expect.


On disk	167 GB
MacBook Pro (M1 Pro, 32 GB, USB4 enclosure)	≈ 5 s / token at a 10.6 GB budget, and it still runs at 2 GB
iPhone 16 Pro (USB 3 port, powered dock)	≈ 17 s / token under 2 GB of memory

Speeds change with the drive, the cable and the budget; the enclosure and cable matter more than the SSD — USB4 against a 10 Gb/s hub is about 3× on the same bytes — and the iPhone needs a powered dock.

Requirements. An Apple-silicon Mac on macOS 15 or later; an iPhone 15 Pro or later on iOS 18 or later; an external NVMe drive with room for the model.

minirun.dev · Docs · This model · GitHub

Provenance


Source model	`deepseek-ai/DeepSeek-V4-Flash-0731`
Source revision	`7872f01b1d1fe23eabc4c98b48bffcef5a386062`
Relationship	byte-preserving repack (no requantization)

License

This repository redistributes model weights owned by DeepSeek under the MIT License, reproduced verbatim in `LICENSE` and copied unmodified from the source repository at the pinned revision above.

The MIT License permits use, copying, modification and redistribution, including commercially, provided the copyright notice and the permission notice are included in all copies or substantial portions of the Software. The weights are provided "as is", without warranty of any kind. Refer to `LICENSE` for the governing text; the summary above is not a substitute for it.