Reference

Find your memory.
Estimate a model.

Windows and macOS memory checks, model-file requirements, and an LM Studio estimation example.

1 · Find your memory on Windows or macOS

Windows

Open Task Manager → Performance → Memory. Record the installed RAM and the available memory with your normal apps open. Then select your GPU and distinguish dedicated GPU memory from shared system memory.

Schematic · not your device

Memory: Installed / in use / available
Dedicated GPU memory: The graphics card’s own capacity
Shared GPU memory: Uses system RAM; not extra RAM

Names vary by Windows version and hardware. Microsoft’s GPU-memory explanation

macOS

Use Apple menu → About This Mac to identify the chip and installed memory. Apple silicon uses unified memory. In Activity Monitor → Memory, check memory used, memory pressure, and swap with your usual apps open.

Schematic · not a measurement

Physical memory: Installed capacity
Memory pressure: How well memory serves the workload
Swap used: Storage being used to support memory

Free memory alone is not the verdict. An Intel Mac can have a different graphics-memory layout. Apple’s memory guide

Storage is where files live. RAM and VRAM are working memory. Free disk space is not interchangeable with either.

Instructions checked 16 September 2026. If your system differs, use its hardware and memory documentation.

2 · Choose a real file before downloading

Read the publisher’s model card: check the task, variant, license, and required app support.
In your runner’s catalog, identify a supported artifact and its publisher or converter. Check the exact model, precision, and format.
Check required companion files, download size, and available disk space. A download size is not the full memory requirement.
Use the weights calculation to shortlist a candidate. Do not download the guide’s illustrative filename as if it were a verified artifact.

For example, Google’s Gemma 3 12B instruction-tuned model card identifies the source model. A converted file for your runner is a separate artifact whose support must be checked.

3 · Estimate a downloaded model with LM Studio

This example uses LM Studio’s command-line tool, lms. Install LM Studio from its official site and follow its CLI setup instructions. In its catalog, select a model artifact your installed runtime supports and download it. The commands below inspect that local file; they do not download one.

Find the exact local model key.
```
lms ls
```
Copy the identifier for your downloaded file. If the command is unavailable, complete the official CLI setup first.
Estimate with an explicit configuration.
```
lms load --estimate-only "MODEL_KEY" --context-length 4096 --gpu off
```
Replace MODEL_KEY with the key you copied. This is a CPU-only example with a 4,096-token context setting, not a recommended setting for every task. For a GPU configuration, use a supported GPU setting and rerun the estimate.
Read the estimate before loading.
Annotated output fields · no invented result
Estimated GPU Memory
Compare against the GPU-accessible pool for this configuration.
Estimated Total Memory
The runner’s overall model-runtime estimate—not your operating system and every other app.
Loadability message
A resource check, not a promise of useful speed or correct answers.
Field names can vary by runtime/version. Do not add total and GPU figures together. For unified memory, they draw on the same installed capacity. Re-estimate after changing context, offload, or the model file.

If the estimate leaves adequate room, load the model with the same configuration in LM Studio and try your task while observing memory pressure and responsiveness. If it fails, reduce the workload or choose a smaller supported model.

Checked 16 September 2026 against LM Studio’s load and estimate documentation. Other runners may not offer the same feature. If yours has no estimate, use its documented requirements and start with a conservative, small supported model while observing memory use.

← Back to the model comparison