FREE · PRIVATE · NO SIGN-UP

Local AI tools for decisions you can trust.

Plan memory, find models that fit and compare options. Your hardware choices stay in your browser.

Check memory ↓Compare models ↓
RAM + MODEL SIZE CALCULATOR

Will this model fit in memory?

Estimate the memory needed for model weights, runtime overhead and operating-system headroom. This is a safe starting estimate, not a speed promise.

SAFE TO TEST8 GB

Estimated safe system RAM floor

~4.5 GB model weights ~1.5 GB runtime allowance 2 GB OS headroom

The estimate uses parameter count and quantization as a planning shortcut. Real files, context length and runtime memory vary. Confirm the exact model file before installing.

HARDWARE SHORTLIST

Models that pass your RAM limit

LocalLens excludes models above its conservative memory floor. Choose a task on the main adviser before treating any result as a recommendation.

22 profiles fit 16 GB RAMRun the full compatibility check →
Qwen 3 14BAlibaba Cloud · General · Coding · Reasoning14 GB floorDeepSeek R1 14BDeepSeek · Reasoning · Math · Research14 GB floorQwen 2.5 Coder 14BAlibaba Cloud · Coding · Tools14 GB floorGemma 3 12BGoogle · General · Writing · Vision13 GB floorLlama 3.2 Vision 11BMeta · Vision · Documents · General13 GB floorQwen 3 8BAlibaba Cloud · General · Coding · Reasoning9 GB floorDeepSeek R1 8BDeepSeek · Reasoning · Math · Research9 GB floorLlama 3.1 8BMeta · General · Writing · Tools9 GB floor
MODEL COMPARISON

Compare two proven profiles

Compare memory, size, inputs, tasks and licence. “Larger” does not automatically mean “better for your job.”

VS
Meta

Llama 3.2 1B

Mature small models for summarization, rewriting and personal information tasks.

Safe RAM
4 GB
Download
~1.3 GB
Inputs
Text
Licence
Llama 3.2 Community
Open verified profile →
Alibaba Cloud

Qwen 3 4B

A widely used hybrid reasoning family with tool support and multilingual strength.

Safe RAM
6 GB
Download
~2.6 GB
Inputs
Text
Licence
Apache 2.0
Open verified profile →
HOW TO USE THESE RESULTS

Fit first. Benchmark second.

Memory fit is only the first test

A model can load and still respond too slowly for your work.

Context uses more memory

Long conversations and large documents increase memory use.

GPU support depends on your engine

VRAM alone does not prove that a runtime supports your exact graphics hardware.