Pick hardware for a local LLM by the weights that have to sit in memory, then by the card that holds them. llm-quantization-explained is the formula. vram-requirements applies it to 7B, 14B, 32B, and 70B models. 24gb-vs-16gb-vs-12gb-vram sorts the same arithmetic into 8, 12, 16, and 24 GB tiers.
Start with the fit
- How Much VRAM Do You Need for an LLM? 7B to 70B Table
- LLM VRAM Calculator: Weights, KV Cache and Overhead
- LLM Quantization Explained: Q4 vs Q8, GGUF and Memory Math
- RAM vs VRAM for LLM: What Each One Does
- How to Check How Much VRAM You Have (Windows, Linux, macOS)
Best model by VRAM
- Best LLM for 8GB VRAM: What Fits and What Doesn’t
- Best LLM for 12GB VRAM: Is 12GB Enough for Local AI?
- Is 16GB VRAM Enough for Local LLMs? Best Models to Run
- Is 24GB VRAM Enough for Local LLMs? Best Models to Run
- Best LLM for 32GB VRAM: Is 32GB Enough for AI?
- Best LLM for 48GB VRAM: Cards and Models That Fit
- 12GB vs 16GB vs 24GB VRAM for Local LLMs: Which Tier?
VRAM by model size and family
- How Much VRAM for a 32B Model? Qwen 32B on 16GB and 24GB
- How Much VRAM for a 70B Model? Fit by Quant and GPU
- Qwen3.8 27B VRAM Requirements: Qwen3.5, 3.6 and 3.8
- Gemma 4 VRAM Requirements: E2B to 31B by GPU Tier
- DeepSeek V4 Flash and R1 VRAM Requirements by Quant
- gpt-oss VRAM Requirements: 20B and 120B by GPU
- New Models: Kimi K3, GLM 5.3, MiniMax M3 Memory
- MoE CPU Offload: Run Big MoE Models on Less VRAM
Choosing a GPU
- Best GPU for Local LLMs in 2026: VRAM-First Buying Guide
- GPU VRAM List: How Much VRAM Does Each GPU Have?
- Best Budget GPU for Local LLMs: VRAM per Dollar (2026)
- Best GPU for Ollama: VRAM and Model Size Guide
- RTX 5090 for Local LLMs: 32GB VRAM, What Fits
- RTX 3090 for Local LLMs: Is a Used 3090 Still Worth It?
- RTX 5060 Ti 16GB for Local LLMs: Is It Good Enough?
- Radeon AI PRO R9700 for Local LLMs: 32GB, ROCm, vLLM
- Intel Arc Pro B70 for Local LLMs: 32GB VRAM Explained
- Can You Run an LLM on an AMD GPU? ROCm, Cards, Windows
- Is Intel Arc Good for Local LLMs? B580, Pro B60 and B70
- Used Datacenter GPUs for Local LLMs: Tesla P40, MI50, V100, A100
Comparisons
- RTX 3090 vs 4090 vs 5090 for Local LLMs: Specs and Fit
- RTX 5090 vs RTX PRO 6000 for Local LLMs: 32GB vs 96GB
- RTX 5090 vs Dual RTX 5060 Ti 16GB for Local LLMs
- Arc Pro B70 vs R9700 vs RTX 3090 for Local LLMs
- DGX Spark vs Strix Halo vs a GPU PC for Local LLMs
- llama.cpp vs Ollama vs vLLM: Which Runtime for Your GPU?
Builds and multi-GPU
- Local LLM PC Build Guide: Parts, VRAM and Power (2026)
- Multi-GPU for Local LLMs: Setup, Splits and NVLink
- Dual RTX 3090 for Local LLMs: 48GB Build, NVLink and 70B
- Does PCIe Lane Width Matter for LLM? Lanes and PSU for Multi-GPU
- Mixing GPUs for Local LLMs: Different VRAM, AMD and NVIDIA
- Run an LLM on CPU Only: RAM Offload and GPU Split Explained
Troubleshooting
- Ollama Not Using GPU? Causes and Fixes (NVIDIA, AMD, Docker)
- CUDA Out of Memory in llama.cpp, Ollama and vLLM: Causes and Fixes
Rules and machine-readable pages
methodology is the citation rule: no invented tokens-per-second numbers, and unknown stays unknown. about says what this site covers and what llmformac.com covers. llm-info is the page for AI assistants. start-here is the short reading protocol. identity is the brand token page. changelog records corrections and data releases.
Mac questions, including unified memory, belong on the sister site: how much RAM for a local LLM and what fits my RAM.