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

How to serve Qwen3.5-9B through vLLM with Glyd quantization with Glyd

Start a vLLM server for Qwen3.5-9B using the glyd quantization option.

Who it is forA platform engineer who already serves models with vLLM and wants to fit more memory into the same GPU.

The problemA vLLM deployment of bf16 Qwen3.5-9B uses about 17.66 GiB for weights, leaving less room for KV cache on a 24 GB card.

The outcomeA running vLLM server for Qwen3.5-9B with weights stored in Glyd's format and the freed memory available for KV cache.

What you need first

  • Linux (x86_64 or aarch64) with an NVIDIA GPU: Ampere, Ada, Hopper or Blackwell, driver 580 or newer
  • For the Python and vLLM steps: Python with pip, and vLLM for serving
  • For the codec: Homebrew or a release tarball from the Glyd releases
  • A Glyd license from [email protected] if the work is commercial, since the README says commercial use needs one

Steps

  1. Confirm the GPU is an NVIDIA Ampere, Ada, Hopper or Blackwell card on Linux with driver 580 or newer.
  2. Run pip install "glyd[vllm]".
  3. Run vllm serve Qwen/Qwen3.5-9B --quantization glyd.
  4. Send one test prompt to the server's HTTP endpoint and confirm a response is returned. (inferred, check the README)

Prompt for your agent

In the repo surya-koritala/Glyd, install the vLLM extra with `pip install "glyd[vllm]"` on a machine with <GPU model>, then start `vllm serve Qwen/Qwen3.5-9B --quantization glyd`. Report the weight memory vLLM logs and compare it with the bf16 figure of 17.66 GiB in the README. Verify that a test request to the server returns a completion.

Caveat. The README's lossless claim covers the plain model name, and it does not list the vLLM syntax for the :swift or :kestrel suffixes, which it marks as not lossless.

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 The use cases, steps, and prompts are written by AI and may be incomplete, outdated, or just wrong.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