Can I run this LLM locally?

Pick an open-weight model, a quantization and a context length. You get the memory it needs and the cheapest GPU, Mac or AI mini PC that fits, priced from real offers. 42 models, 50 devices, 44 with a current price.

Memory needed at Q4_K_M with 32K context

Estimated total (weights, KV cache and runtime overhead) and the cheapest single device in our catalog with enough usable memory (legacy server cards excluded).

FAQ

How much VRAM do I need to run a 70B model?

At Q4_K_M with a 32K context, Llama 3.3 70B needs about 56.2 GB by our estimate (43.2 GB of weights plus KV cache and overhead). That is more than any single consumer GPU. The cheapest setup in our catalog that fits is 4× NVIDIA Tesla V100 16GB (PCIe or SXM2), from about €1,099. A shorter context or the Q8_0 KV cache lowers the requirement.

Why does context length matter?

Every token in the context keeps keys and values in the KV cache. For models with many full-attention layers this can add tens of gigabytes at 128K tokens. Hybrid and sliding-window models (Qwen3.5+, Gemma 4, gpt-oss, Nemotron 3) need far less.

Where do the prices come from?

From AI Hardware Deals (hardware.branolabs.com), which tracks real offers from retailers and marketplaces with anti-scam filtering. We show the cheapest current offer per device, converted to EUR at approximate rates.