AIToolScan

whichllm Overview

whichllm finds the local LLM that actually runs and performs best on your hardware. It ranks models by real, recency-aware benchmarks — not parameter count — and lets you run a pick instantly with one command.

• Hardware Auto-Detect: Detects NVIDIA, AMD, Intel, Apple Silicon, and CPU-only setups, then ranks HuggingFace models that fit your VRAM\/RAM.
• Benchmark-Aware Ranking: Merges LiveBench, Artificial Analysis, Aider, Chatbot Arena ELO, Open LLM Leaderboard and more — with confidence tags and stale-score demotion.
• One-Command Start: Run uvx whichllm@latest with no project setup; also install via pip, brew, or uv tool install whichllm.
• GPU Simulation & Planning: Pretend you have an RTX 4090\/5090, simulate multi-GPU, compare upgrades, or reverse-plan GPUs with whichllm plan.
• Safer LM Studio-Style Picks: Use --gpu-only, --speed usable, and --vram-headroom to keep full-VRAM, comfortable-speed recommendations.
• Run & Snippets:whichllm run downloads and chats instantly; whichllm snippet prints copy-paste Python for any model.
• Scriptable Output: Markdown tables for Slack\/Discord and --json for pipelines — no TUI required.
• Open Source: MIT-licensed Python 3.11+ CLI by Andyyyy64, with live HuggingFace data and curated offline fallbacks.