syv-ai / HyperQwen
Serve large Qwen models fast on the GPUs you actually own. Qwen3.8-27B on a single 24 GB card with vLLM: 127 tok/s single-user (381 when the answer quotes the prompt), ~1,035 tok/s at 64 concurrent, 150k-262k context. vLLM patches, requant pipeline, benchmarks.
Why this job: topics: local-llm, llm-inference, vllm, quantization
What it does
First start pulls the image (9.5 GB) and requantizes the model (~20 GB, once, into ./models), then serves on :18020. One GPU runs one mode at a time.
Get started
No single install command. Read the steps in the README (opens in new tab)
- Maintenance
- Active
- Last commit 6 days ago
- Stars
- 2,626
- +136 this week
- Licence
- Apache-2.0
- Allows business use
- Downloads for
- None listed
- No release files
- Latest version
- None yet
- No releases
- Built with
- Python
- 58% of the code
Is it looked after?
- Contributors
- 4
- Last 90 days
- Issues
- 99% closed
- 1 still open
- Pull requests
- 264 merged
- 5 waiting
- Releases
- None
- Never published
- No contributing guide
- No security policy
- No code of conduct
Similar repos
| Repo | Stars | Last commit | Licence | Downloads for |
|---|---|---|---|---|
| JakeATX / llamAmpere | 299 | today | MIT | No downloads |
| Sandermage / sndr_core_engine | 133 | 4 months ago | Apache-2.0 | No downloads |
| LMCache / LMCache | 12,310 | today | Apache-2.0 | Linux |
More: languages, topics and project record
Languages
- Python 58%
- Shell 17%
- Cuda 11%
- C++ 10%
Topics
- consumer-gpu
- cuda
- inference-optimization
- kv-cache
- llm-inference
- local-llm
- quantization
- qwen
- qwen3
- rtx-3090
- speculative-decoding
- vllm
Project record
- JobLocal LLMs
- StatusListed
- Commits694
- Forks338
- Watching18
- Stars in 28 days+664
- CreatedAug 15, 2026
- First seen by usNov 7, 2026
- CheckedNov 7, 2026