Free VRAM calculator and guides for choosing a GPU or Mac to run open-weight large language models locally with Ollama, llama.cpp or LM Studio.
About LLMHardware.io
LLMHardware.io helps people choose hardware for running open-weight large language models (LLMs) on their own computer instead of a cloud API. The VRAM calculator is free with no signup. Enter a model's size in billions of parameters, a quantization level, the context length and the key-value (KV) cache type, and it estimates the GPU or unified memory needed and shows the hardware that fits. Guides cover GPUs, Apple Silicon Macs, mini PCs, memory needs for model families such as Qwen3, Llama and Gemma, quantization, and setup with Ollama, llama.cpp and LM Studio. Estimates come from a published formula, spec sheets and model cards; the site does not measure hardware. AI helps write it.
PlatformsWeb
Key features
- VRAM calculator for any model size at Q4_K_M, Q5_K_M, Q8 or FP16, with context lengths up to 128K and a KV-cache type setting
- Mixture-of-Experts option that estimates a GPU and system RAM split for llama.cpp hybrid offload
- Can I Run It? checker with a Q4, Q8 and FP16 verdict for one GPU or Mac and a shareable link
- Guides on GPUs, Apple Silicon, mini PCs, model memory needs, quantization, and Ollama, llama.cpp and LM Studio setup
- Embeddable calculator snippet for blogs and documentation sites, no signup required
Made by
LLM Hardware team
🏷️ Get an embeddable badge for LLMHardware.io
Show it off on your own site — updates automatically, links back here.
Reviews
No reviews yet — be the first to share your experience.
Similar tools
FitGPU
Free local LLM hardware picker: VRAM estimates, GPU compare, real Ollama tok/s benchmarks
LLM Memory (VRAM) Estimator
Free browser tool: how much GPU memory (VRAM) a local LLM needs for its size, quantization and context length
Modal
Serverless compute for AI teams
Runpod
Cloud GPUs for AI, on demand
Beam
Serverless GPU infrastructure for AI
Cerebrium
Serverless infrastructure for ML