LM Studio
Desktop app that runs open models on your own machine, now wrapped in Bionic, an agent that takes multi-step work off your hands locally.

Evaluation
What to know
Editorial notes based on the vendor's own documentation and public pricing, plus hands-on testing where a score or review is shown above. They describe fit for a job, not absolute quality.
Best for
- People who want an AI agent without sending their documents to a vendor
- Developers and analysts with a workstation GPU and privacy constraints
Not ideal for
- Users on low-end laptops expecting frontier-model quality
- Teams that need central administration and shared audit logs
Key features
- Desktop app that downloads and runs open models locally
- Bionic agent for multi-step work, using local models first
- Cloud credits on paid plans for jobs that exceed local hardware
- Model-agnostic setup across local and hosted models
- No account required for local-only use
Pros
- Sensitive work can stay on the machine, which hosted assistants cannot promise
- Free tier is genuinely usable, including the agent
- Switching models is a few clicks, so you can match cost to task
Cons
- Agent workloads want a strong GPU; a thin laptop will hit limits fast
- Cloud credits and plan caps make heavy use a budgeting exercise
- Local models still trail frontier hosted models on hard reasoning
ToolifyNav verdict
LM Studio has grown from a model playground into a local agent runner, and that shift is the story: the interesting question is no longer which model you can load but what an agent can finish on your own hardware. For anyone handling client files, unpublished manuscripts or internal documents, running the work locally removes an entire category of risk. The limits are physical — memory, GPU time, and how much reasoning a local model can sustain — so it rewards people who plan their workload. Teams that need shared governance should treat it as a personal tool, not a company platform.
About
About LM Studio
LM Studio began as the easiest way to download a model and chat with it on a laptop. The current version pushes in a different direction: Bionic is an agent that runs on your machine, plans multi-step tasks, uses local models by default and can reach for cloud capacity when a job outgrows the hardware. The pitch on the site is ordinary desk work — research, drafting, code — done by an agent rather than a chat window you prompt line by line.
Two things make it interesting for people who keep sensitive material around. The default is local: models download and run on your own GPU, and the agent's work stays on the machine unless you connect something. And it is model-agnostic, so the same agent can sit behind a small quantised model for quick tasks and a larger hosted one for heavy reasoning.
The free tier includes the agent and local models; paid plans at $20 and $100 a month add cloud credits, and bulk token purchases are offered for heavier use. The honest caveat is hardware: a decent agent workload still wants a capable GPU, and everything the agent does through integrations deserves a permissions review.
Related
More in Chatbots & Virtual Companions
Same category — not a curated competitor list.
Discover More
Weekly digest
Get weekly AI tools and hands-on reviews
A short digest of new tools and hands-on breakdowns. Unsubscribe anytime.


