Local LLM memory calculator
Budget weights, quantization, context cache and system headroom. Share every assumption or export the result as CSV.
Plan your memory →LocalClaw / Tools & references
Free resources for choosing, running and explaining local AI. Start with a calculation, follow the sources and share something others can reproduce.
Budget weights, quantization, context cache and system headroom. Share every assumption or export the result as CSV.
Plan your memory →Explore language, speech, image, video, 3D, music and vision records. Check each model’s local path and hardware notes.
Find a model →Compare memory tiers and understand the difference between system RAM, dedicated VRAM and Apple Silicon unified memory.
Explore hardware →Try supported small models in your browser. Check each demo’s download, device and browser requirements before running it.
Explore browser demos →Start from a practical workflow and follow the linked model, software and hardware requirements.
Find a project →Historical snapshot dated June 19, 2026: 191 LLM records, 58 speech records and 15 machines. Its original CC BY 4.0 terms apply. It is not the current catalogue.
Open the historical dataset ↗For writers, educators & builders
Link directly to the tool or model page you used. For memory estimates, include the calculation link, method version and inputs. For model compatibility, cite the individual model page and its original sources, and distinguish a published hardware floor from a measured result.
LocalClaw. Local LLM Memory Calculator, method v1.0 (November 2, 2026). https://localclaw.io/local-llm-memory-calculator. Planning estimate, with input settings supplied in the result link.
The historical GitHub release retains its published CC BY 4.0 licence. It includes hosted-model references and estimates that must be checked before current use. New LocalClaw compilations and compatibility analyses have separate data reuse terms; the historical licence does not extend to them. Model weights and third-party software keep their own licences.
To report a calculator issue, open an issue with the result link and expected behaviour. For dataset reuse questions, contact LocalClaw.