SOURCE-CHECKED EDITORIAL GUIDE · #10
LM Studio
Independent LM Studio guide covering local-model setup, hardware fit, offline expectations, licenses and practical evaluation steps.
Category: AI Platform
LM Studio is a desktop application for discovering and running compatible language models locally, and it can expose a local server for applications. Local execution can reduce the data sent to a hosted model, but privacy still depends on downloads, updates, connected tools and the user's own configuration.
Open official website Browse all guides
At a glance
Primary access
Windows, macOS and Linux desktop app
Plan and availability
LM Studio features and limits can vary by account, region and plan; confirm the current provider page before relying on a feature.
Evidence standard
This LM Studio guide uses first-party product, help and legal sources; it does not use a paid ranking or user-rating score.
Sources reviewed
What matters before you choose
Choose a model that fits available memory and verify its license before business use. Test latency, output quality and resource use on the actual computer rather than relying on model size alone. A local model can still produce inaccurate content, so factual and code outputs need the same verification applied to cloud systems.
Decision questions
- Does the chosen model fit the actual RAM, GPU memory and storage available?
- Does the model license permit the intended use?
- Which parts still contact the internet when the network is disconnected?
A practical test scenario
Record the test computer's operating system, RAM, GPU memory and free storage, then choose a model whose license and size you have checked. Measure first-token delay, sustained speed, memory use and answer quality on three representative prompts. Disconnect the network and repeat the essential workflow to learn which parts are genuinely local and which still depend on online services.
Stop and reassess if
- The model does not fit the available memory or makes the machine unusable for the intended workflow.
- The model license is unclear for the planned personal or commercial use.
- A supposedly local workflow still needs an unreviewed online connector, download or telemetry path for essential steps.
Run this LM Studio test yourself with the account and device you actually plan to use. AI Tools Galaxy does not claim to have tested every LM Studio plan, model or regional configuration; keep dated notes so you can compare later provider changes.
AI Tools Galaxy evidence status
SOURCE-CHECKED
This page is based on first-party source review and a reader-repeatable evaluation method. No AI Tools Galaxy hands-on benchmark result is currently published for this product.
A repeatable evaluation
For LM Studio, use a harmless sample that represents your real task, then record the exact account, plan, model or mode, enabled connections and test date. Decide what a correct LM Studio result looks like before the run so a polished answer cannot substitute for evidence.
- Record operating system, RAM, GPU memory and available storage.
- Check the model card and license before downloading a model.
- Measure first-token delay, tokens per second and memory use on three real prompts.
- Disconnect the network and verify which parts of the intended workflow still function.
Record the result
In your LM Studio notes, separate factual errors, unsupported claims, privacy concerns, correction time and plan limitations. Reuse the same inputs when comparing an alternative and repeat the test after a material LM Studio model, plan or policy change.
Official sources checked
Editorial disclosure: AI Tools Galaxy independently reviewed the LM Studio first-party sources above. This is not a sponsored placement or a claim of laboratory testing. LM Studio features, plans and terms can change, so confirm the current source before an important decision.
Related source-checked guides
Continue your research
Keep the next lm-studio comparison tied to the same task and risk. Use the closest topic hub or checklist below so you can test lm-studio alternatives against the same evidence standard instead of switching to an unrelated workflow.