Best local LLMs by hardware
One recommendation per memory budget, computed from the same index as the calculator — not a list of everything that fits. Pick your capacity; each page names the model for general use, for coding and for images, with the memory maths and what breaks when you ask for more context.
- Best local LLM for 8GB35 models fit; Stable LM 2 12B Chat at AWQ INT4 (7.0 GB) is the general-purpose pick.6 · RTX 5060 Ti 8G, RTX 5060, RTX 4060 Ti 8G…
- Best local LLM for 12GB46 models fit; DeepSeek-V2-Lite Chat at AWQ INT4 (9.1 GB) is the general-purpose pick.8 · RTX 5070, RTX 4070 Ti, RTX 4070 Super…
- Best local LLM for 16GB52 models fit; Mistral Small 24B Instruct at AWQ INT4 (13.2 GB) is the general-purpose pick.12 · RTX 5080, RTX 5070 Ti, RTX 5060 Ti 16G…
- Best local LLM for 24GB65 models fit; Seed-OSS 36B Instruct at AWQ INT4 (19.9 GB) is the general-purpose pick.3 · RTX 4090, RTX 3090, Radeon RX 7900 XTX
- Best local LLM for 32GB66 models fit; Mixtral 8x7B Instruct at AWQ INT4 (24.9 GB) is the general-purpose pick.2 · RTX 5090, Instinct MI100 32G
- Best local LLM for Apple silicon66 models fit; Mixtral 8x7B Instruct at Q4_K_M (30.1 GB) is the general-purpose pick.18 Apple silicon · Mac M5 Ultra 512G, Mac M5 Ultra 256G, Mac M5 Max 128G…