May 15, 2026
Your PC got rated — chaos followed
Show HN: Find the best local LLM for your hardware, ranked by benchmarks
This tool promises your perfect AI match — commenters instantly started arguing
TLDR: whichllm is a new tool that tries to tell you which AI model is best for your computer based on real test scores, not just size. Commenters loved the idea but immediately fought over broken installs, questionable picks, and whether its hardware math can be trusted.
A new command-line tool called whichllm is pitching itself as the matchmaker for local AI, promising to scan your computer, figure out your graphics card, memory, and processor, then tell you which chatbot model will actually run well on it. In plain English: instead of just saying “this big model fits,” it tries to say “this one is actually the smartest choice for your machine.” Cute idea. Predictably, the comments turned into a live episode of Tech Support: Civil War.
The biggest fight? Can any tool really predict this stuff accurately? One commenter immediately zoomed in on memory math, basically saying that if the estimate is even a little wrong, the whole thing can crash and burn. Another questioned whether the tool is simplifying too much, since different AI models use memory in very different ways. Translation for normal people: the app says it knows what your computer can handle, and the crowd is responding, “Oh yeah? Prove it.”
Then came the classic open-source drive-by: “Brew install is broken.” Ouch. That same commenter also called the recommendations “pretty rubbish,” saying the app suggested older Qwen models even though newer ones were already running fine on their Mac. That was the thread’s main scandal: is this a smart recommendation engine, or is it confidently serving yesterday’s leftovers?
Meanwhile, the peanut gallery added flavor with requests for more model formats, a homemade sizing rule, and the timeless Hacker News subtweet: “What’s new regarding llmfit?” In other words, even when a tool tries to end AI confusion, the community’s favorite hobby remains making it dramatically more confusing
Key Points
- •whichllm is a CLI tool that detects hardware and ranks local LLMs from Hugging Face based on fit, benchmark quality, and estimated speed.
- •The article emphasizes that the tool does not simply recommend the largest model that fits in VRAM; it uses benchmark evidence and model recency to rank options.
- •Sample recommendations are provided for RTX 5090, RTX 4090/3090, RTX 4060, Apple M3 Max, and CPU-only systems, with example token-per-second estimates.
- •The ranking system merges benchmark sources including LiveBench, Artificial Analysis, Aider, Chatbot Arena ELO, and the Open LLM Leaderboard, and discounts lower-confidence score sources.
- •whichllm supports features including hardware auto-detection, GPU simulation, task profiles, JSON output, direct model execution, Python snippet generation, and GGUF support via llama-cpp-python.