DEDUPED Hot AI repos, turned into use cases Get 5 a day

How to run the experimental 125B Flash model from disk with Pooled

Load Qwen3.8 Flash, which keeps most of its experts on disk, in one Chrome tab and speed it up by joining a second PC.

Who it is forAn experienced user with a large SSD and a capable GPU who wants to test a 125B mixture-of-experts model that is much larger than the GPU memory of most computers.

The problemA 125B model does not fit in typical GPU memory, so running it normally needs far more hardware than a single personal computer.

The outcomeThe model runs with its experts stored on disk and cached on the GPU as needed, and a second computer joined with the command line lends memory to make it faster.

What you need first

  • A browser with WebGPU support, such as Chrome, on a computer or phone with a GPU
  • A room link from someone who started a room, or the Start a room button on the site
  • Free disk space and a stable connection for the model download, about 17 GB for the 27B model and about 22.5 GB for the 35B MoE model
  • Node 22 or newer for the command line bridge (npx @pooled/cli)

Steps

  1. Open a room in Chrome on the computer that has the SSD space, and pick Qwen3.8 Flash in the room's model picker.
  2. Accept the prompt to store the first download, which is about 99 GB, on the device.
  3. On a second PC, run npx @pooled/cli join followed by the room link or code, as described in cli/README.md. (inferred, check the README)
  4. Send a short question and compare the decode speed to the figures in the README.

Prompt for your agent

In the Nehanth/pooled repository, open a room in Chrome, select Qwen3.8 Flash, accept the roughly 99 GB storage prompt, and send 'Hi'. Expected output: a reply after a few seconds on a single machine. Verify that the room shows the model as loaded and the reply uses the 64K context setting.

Caveat. The README marks this model as experimental, with context up to 64K, slower decode than the other models, and loading and requirements that may change between releases.

Share: X LinkedIn Hacker News Email Copy link

Disclaimer

Deduped lists open-source repositories that belong to other people. We do not own, maintain, audit, or vouch for any of them, and a repo appearing here is not an endorsement or a safety check. The use cases, steps, and prompts are written by AI and may be incomplete, outdated, or just wrong. Where a use case is marked "Seen in the wild", we link to public posts where someone described something similar, but we have not verified those posts and the use case itself is still our AI's own write-up. Use all of this at your own risk. Only run code, commands, or agents on systems and accounts you are authorized to use, and read the code and every command before you run it, especially anything that touches credentials, files, or the network. Nothing here is professional, security, legal, or financial advice. All trademarks and project names belong to their respective owners, and their use here does not imply affiliation. Content is provided as is, without warranties of any kind. Found a mistake, a safety problem, or something you want removed? Email [email protected] and we will look at it promptly. Full disclaimer page