whichllm ranks best local LLMs for hardware

whichllm solves the "black box" problem of local LLM performance by providing empirical rankings based on a user's specific GPU and RAM.

• –Hardware-specific scoring accounts for VRAM overhead, memory bandwidth, and model architecture for accurate performance predictions.
• –The "plan" command offers a reverse-lookup for hardware buyers, identifying the components needed to run specific models like Llama 3 or Qwen.
• –Integrated execution via isolated uv environments makes it a one-command alternative to complex setups for rapid model exploration.
• –v0.5.2 improves Apple Silicon estimation and multimodal model scoring, ensuring unified memory and vision-capable LLMs are correctly ranked.
• –Live data fetching from HuggingFace ensures rankings track the latest releases and benchmark shifts in real-time.