Local AI Stats.

Local LLM Rankings November 2026

Local LLM rankings.Compare benchmark scores and model size to find a practical local model.

Skip to model comparison

The Pareto Frontier of Local LLMs

Up to 80B parameters or 40 GB of model weights · Runtime memory needs extra space.

025507510005101520LFM2.5 230M: 20.25 MMLU-Pro, 0.15 GBMiniCPM5 2B: 70.8 MMLU-Pro, 1.56 GBQwen3.5-4B: 79.1 MMLU-Pro, 2.74 GBQwen3.5-9B: 82.5 MMLU-Pro, 5.68 GBQwen3.5-27B: 86.1 MMLU-Pro, 16.74 GBQwen3.6-27B: 86.2 MMLU-Pro, 16.82 GBQwen3.6-27BQwen3.5-27BQwen3.5-9BQwen3.5-4BMiniCPM5 2BLFM2.5 230MMMLU-ProGb
BENCHMARK

Knowledge and reasoning across many subjects.

MODEL SIZE
Best trade-offs
Best fit: Qwen3.5-4B, 79.1 on MMLU-Pro.

PARETO CHAMPION (MMLU-Pro)

Qwen3.5-4B

Qwen · Q4_K_M

79.1/100Test score
2.74GBMemory footprint
28.86Score / GB
Get model weights View full benchmark record

COMPARE MEMORY OPTIONS

Computers for Qwen3.5-4B

These configurations pass a weight-size check for the 2.74 GB file. Compare the platforms, then check the exact Amazon listing.

Change platform or budget ↗

Weight size is not tested compatibility or speed. Check runtime support and memory for context before buying. Prices are dated references. As an Amazon Associate I earn from qualifying purchases.

How these options are selected

Up to three platforms, with the lowest dated reference price on each. Unknown prices follow known prices. Each memory pool keeps a reserve: 30% (minimum 2 GB) for system or unified memory; 20% (minimum 1 GB) for dedicated VRAM. System RAM is for CPU inference. This check does not establish GPU allocation, software support or context capacity.