Your saved laptops. Pick from these to compare.

Best laptops for running local LLMs. Strict 4-bit quant model: we only count memory the GPU can actually use fast — VRAM on discrete GPUs (model must fit), 75% of unified RAM on Apple Silicon, or system RAM on iGPU (slow). No fake "with offload" bonus. Number shown is the largest model (B params) the laptop can run with ~1.5 GB context overhead.

Pick a use-case profile and we'll re-weight every stat to find your top matches across all 290+ laptops.

All laptops ranked by overall performance score.

Pick a laptop, hit Check Value. We compute an expected price from its specs (GPU, CPU, VRAM, RAM, AI, battery, build) and compare it to peer laptops in the same performance tier. Shows you whether it's a steal, fair, or overpriced — with the math.

Laptops ranked by performance-per-dollar. Pick a metric — score per $1000, GPU per $1000, AI capability per $1000, or battery hours per $100 — and see which laptops give you the most bang for your buck.

Visual breakdowns of the dataset. The scatter plot shows price vs overall score — laptops below the trend line are the best deals (Pareto winners are circled). The histograms reveal where the market clusters and where the gaps are.

Answer a few quick questions and we'll auto-build a custom recommendation profile from your real priorities — less abstract than picking a preset.

Pick a specific local LLM and we'll surface the cheapest laptops that can actually run it at usable speed. Inverts the usual question — instead of "what can this laptop run?", ask "what's the cheapest hardware that runs this model?"

Build your own machine. Performance scores auto-derive from real specs.

Spotted a bug, want a new feature, or just want to say hi? Two ways to reach me.

Keyboard shortcuts, how to use each feature, and FAQ.

How the scores, tiers, and predictions on this site are computed — sources, formulas, and known limitations.

Side-by-side specs and performance scores. Best value per row in green, worst in red.