C++ · MIT
kekzl/imp
Run large AI chat models privately on your own NVIDIA RTX 50 or RTX PRO graphics card. On one RTX 5090, Qwen3.8-27B writes about 70 words per second, 1.4x faster than vLLM. Even a model almost four times larger than the card's memory runs. Works with apps made for ChatGPT or Claude. Starts with one docker command.
★ 124
Stars
6 forks · 3 open issues
flat
Growth · measured
Stars counted between two of our own snapshots roughly a week apart. The repository page gives the exact span.
21.1
Momentum / 100
see the breakdown below
8mo ago
Created
last push 2d ago
Star history
How this score was built
| Component | Weight | Earned | Why |
|---|---|---|---|
| Velocity | 55 | 0.0 | flat, log-scaled against a 750/day ceiling |
| Acceleration | 15 | 0.0 | needs 7-day and 30-day rates side by side |
| Freshness | 15 | 15.0 | last push 2d ago; decays to zero at 180 days |
| Scale | 15 | 6.1 | 124 stars, log-scaled against 150k — stops a 3-star repo topping the board |
| Momentum | 100 | 21.1 | sum of the above |
Detail
| Topics | ai-agentsblackwellcudainference-enginellm-inferencelocal-llmmixture-of-expertsnvfp4openai-apiopenai-compatiblequantizationqwenrtx-50-seriesrtx-5090rtx-pro-6000safetensorsself-hostedtool-callingvllm |
| Matched on | LLM & Inference (llm-inference) |
| Short window | +0 stars over 7 days = 0.0/day |
| Long window | no snapshot 14–45 days old yet |
| Lifetime average | 0.49 stars/day over 255 days |
| First seen here | 2026-10-29 |
| Last read | 2d ago |