AutoYou
Freemium
Users wanting fully private local AI
LM Studio
FreeAPISelf-Hosted
LM Studio is a free desktop app for downloading and chatting with local models. It also ships a headless server, llmster, plus local APIs and Python and TypeScript SDKs for using those models from your own code.
Free to run models locally; optional cloud plans: Bionic+ $20/mo, Pro $100/mo · Checked Oct 16, 2026
Similar tools
llama.cppFreellama.cpp is the MIT-licensed C/C++ engine that runs GGUF language models on your own hardware…
AutoYouFreemiumAutoYou runs a local AI server on your Windows, macOS, or Linux machine, then lets your phone…
LM-Kit OneFreemiumThe vendor describes LM-Kit One as a self-hosted AI application server that exposes OpenAI…
Atlas Inference EngineFreeThe vendor page benchmarks Atlas at 3.1x the decode throughput of vLLM on Nvidia DGX Spark…
Screenshots 1
Summary
We tested the headless server, llmster 0.0.25-1, in Docker on a CPU-only server with the same 3B model (qwen2.5:3b) we used for Ollama. Installing took one script and under a minute. The model itself loaded in 8.1 seconds, but the lms load command took 44.6 seconds on a cold start: about 34 seconds passed before the engine even started, and the logs don't say what it was waiting on. It also tried to reach three lmstudio.ai hosts. We blocked them and it kept serving, so it works offline once your models are downloaded. On the same model and cores, Ollama generated text faster with short and medium prompts; with our longest prompt they were even. The app and llmster are closed source and free under LM Studio's terms; only the lms command-line tool is MIT-licensed. We did not test the desktop app or GPU speed.
Bottom line: Pick LM Studio if you want a desktop app for local models with an API on the side. If you only need a headless server, Ollama started faster and generated text as fast or faster in our test.
Pricing Plans
Free|SubscriptionLast verified 3 weeks ago- Price
- Free to run models locally; optional cloud plans: Bionic+ $20/mo, Pro $100/mo
Free
Free
Run local LLMs on your machine.
- Run local LLMs using llama.cpp and MLX
- Bionic Agent
- Offline voice transcription
- LM Link for up to 5 devices
- Limited web search
Bionic+
$20per month
Optional US-hosted cloud inference for open-source models, zero data retention.
- Everything in Free
- US-hosted open source models (Kimi K3, GLM 5.3, DeepSeek V4 Flash and more)
- Discounted bulk tokens
- Web search and page extraction
Pro
$100per month
More included cloud usage.
- Everything in Bionic+
- 5x usage limits
- Discounted bulk tokens
- Early access to new features
View full pricing on lmstudio.ai →
Pricing may have changed since last verified. Check the official site for current plans.
We installed and ran it
- How we installed it
- Vendor one-line install script (curl | bash)
- Install script run time
- 53.9 s
- Docker image size on disk
- 4.68 GB
- Model we used
- qwen2.5:3b (Q4_K_M GGUF, 1.9 GB)
- Model load time from its engine's log, model file not cached
- 8.1 s
- lms load command, model file not cached
- 44.6 s
- Memory used by the model process
- 2 GB
- Outside hosts it tried to reach (we blocked them)
- extensions.lmstudio.ai, lmstudio.ai, versions-prod.lmstudio.ai
- The app and its headless server (llmster) are closed source and free under LM Studio's terms; the lms command-line tool is MIT.
- It tried to reach lmstudio.ai, versions-prod.lmstudio.ai and extensions.lmstudio.ai. We blocked all of them and it still loaded and served the model.
- The lms load command took 44.6 s on a cold start, but its llama.cpp engine needed only 8.1 s of that to load the model: about 34 s (33 to 36 s) passed between the load request and the engine starting. Its log does not show what it waited on.
Measured against Ollama
| Measured | LM Studio | Ollama |
|---|---|---|
| Model load time from the runtime's own log, model file not cached (s) | 8.1 | 5.9 |
| Start-up, not the same measurement: Ollama = first answer from a fresh server; LM Studio = lms load command (s) | 44.6 | 6.7 |
| Text generation speed, 53-token prompt, relative | 1x | 1.38x |
| Text generation speed, 441-token prompt, relative | 1x | 1.22x |
| Text generation speed, 723-token prompt, relative | 1x | 1x |
| Prompt reading speed, 53-token prompt, relative | 1x | 1.32x |
| Prompt reading speed, 441-token prompt, relative | 1x | 1.77x |
| Prompt reading speed, 723-token prompt, relative | 1x | 1.58x |
| Memory used by the model process (GB) | 2 | 2.2 |
| Docker image size on disk (GB) | 4.68 | 9.37 |
Same GGUF model file, same 3 cores, measured back to back in one session. Speed is shown for each of our three prompt lengths: Ollama's lead in generation speed shrinks as the prompt gets longer. The two start-up rows are different measurements: Ollama's is the first 1-token answer from a freshly started server (the model loads on that request); LM Studio's is the lms load command, most of which passed before its engine started loading the model.
Test machine: 8-vCPU AMD EPYC 9354P cloud VM, each test limited to 3 cores and 10 GB RAM, no GPU, no internet (isolated Docker network). Times in seconds are from this CPU-only machine; a desktop with a GPU will be much faster. The VM is shared, so text-generation speed is shown only as a ratio between tools run back to back, not as tokens per second. Docker image size is the on-disk size from docker image ls for every tool. Local AI tools we tested · How we test · raw data
Find your fit
Before you choose LM Studio
More options in Local Inference Runtimes. Compare pricing, platforms, and features side by side.
LM-Kit One
Freemium
Companies needing data-residency compliance
Field notes
All field notes →- LM Studio cold starts: the load command and engine timed different things
Our lms load command took 44.6 seconds while the engine reported 8.1 seconds loading the model. The two timers measured different boundaries.
- How much RAM did our 3B local model actually use?
We observed 1.6–2.2 GB in the model processes running one 3B GGUF. Those numbers are not whole-system RAM requirements or a recommendation for a particular laptop.
Pros
- A desktop app and a headless server (llmster), so you can chat with models or serve them to your apps.
- Local APIs plus Python and TypeScript SDKs.
- Works offline: with its update and extension hosts blocked, it still loaded and served the model.
- Used 2 GB of memory with our 3B model, a little less than Ollama's 2.2 GB.
Cons
- Slow cold load: lms load took 44.6 s, about 34 s of it before the engine started, with nothing in the logs to explain the wait.
- Closed source and free under LM Studio's terms; only the lms CLI is open source (MIT).
- Ollama generated text up to 1.38x faster on the same model and cores.
About
- Platforms
- macOS (Intel and Apple Silicon), Windows, Linux (x64 and ARM64)
- API Available
- Yes
- Self-Hosted
- Yes
- Last Updated
- June 9, 2026
Best For
Who it's for
- People who want a desktop app for downloading and chatting with local models
- Developers who want a local API and SDKs alongside the app
What it does well
- Chat with a downloaded model
- Serve a local model to your own app
Integrations
Compare LM Studio
Frequently Asked Questions
- Is LM Studio free?
- LM Studio offers a free option. Hosting, model usage, or optional services may have separate costs; check the pricing details and official terms.
- Does LM Studio have an API?
- Yes. LM Studio exposes a developer API. See the official documentation at https://lmstudio.ai for details.
- Can I self-host LM Studio?
- Yes. LM Studio supports self-hosting on your own infrastructure.
- When was LM Studio released?
- LM Studio was first released in 2023.
- What platforms does LM Studio support?
- LM Studio is available on: macOS (Intel and Apple Silicon), Windows, Linux (x64 and ARM64).
People also compared
Curated lists that include this category
LM Studio is a free desktop app for downloading and chatting with local models. It also ships a headless server, llmster, plus local APIs and Python and TypeScript SDKs for using those models from your own code.
We tested the headless server, llmster 0.0.25-1, in Docker on a CPU-only server with the same 3B model (qwen2.5:3b) we used for Ollama. Installing took one script and under a minute. The model itself loaded in 8.1 seconds, but the lms load command took 44.6 seconds on a cold start: about 34 seconds passed before the engine even started, and the logs don't say what it was waiting on. It also tried to reach three lmstudio.ai hosts. We blocked them and it kept serving, so it works offline once your models are downloaded. On the same model and cores, Ollama generated text faster with short and medium prompts; with our longest prompt they were even. The app and llmster are closed source and free under LM Studio's terms; only the lms command-line tool is MIT-licensed. We did not test the desktop app or GPU speed.
Pick LM Studio if you want a desktop app for local models with an API on the side. If you only need a headless server, Ollama started faster and generated text as fast or faster in our test.
Contribute
- Work on LM Studio? Claim this listing.
- Embed a badge for your README or site.
Sign in to rate this tool or suggest an edit.