LM Studio
LM Studio runs open-source LLMs locally with Bionic, an agent for coding, documents and automation.
If your work touches files you would rather not upload, LM Studio is the local agent worth installing — Canvas, session introspection and Skills make Bionic more than a chat window, and the Free tier covers local inference, voice transcription and five linked devices. The catch is the cloud side: Bionic+ and Pro are credit-based, so sustained heavy use outruns a flat $20 or $100 monthly. Pick it for privacy and agent depth.
Verified 22h ago · liveness 75/100 · cite: rightaichoice.com/tools/lm-studio
- Privacy-conscious developers running agentic AI over sensitive source files and documents
- Engineers who want a local coding assistant behind an OpenAI-compatible endpoint
- Researchers testing open models such as Kimi K3, Muse Glimmer 30B and DeepSeek V4 Flash
- Mac users chasing the fastest local Qwen3.8 path via the Splash engine
- Buyers who need GPT- or Claude-class proprietary models driving the agent
- Teams wanting per-seat subscription management today — organization billing covers credits only
- Anyone unwilling to pick models, manage VRAM and fight the occasional load failure
We scan live Reddit threads, YouTube comments, X posts, G2 reviews and other communities — and hand you an honest verdict in under a minute.
- Honest verdict, not marketing
- Real pros & cons from real users
- Attributed quotes with receipts
3 free scans · no card needed
Skip LM Studio if you need an agent that runs GPT-4- or Claude-class proprietary models, since Bionic is built around open models running locally or through US-hosted inference on Bionic+ at $20/mo or Pro at $100/mo.
Cloud usage past your plan allowance is covered by pay-as-you-go credits, so a heavy month on Bionic+ or Pro bills above the $20/mo or $100/mo subscription.
LM Studio's Free tier is more capable than most local runtimes at $0: full Bionic Agent, local inference, offline voice transcription, and LM Link for five devices. Bionic+ at $20/mo targets individuals who need US-hosted frontier open models with ZDR, and Pro at $100/mo adds 5× usage limits for heavy agent users. Both cloud tiers are month-to-month with no commitment, and credits cover overage. Against cloud-only assistants that charge similar monthly rates for closed models, LM Studio's edge
In short
LM Studio — LM Studio runs open-source LLMs locally with Bionic, an agent for coding, documents and automation. Best for Privacy-conscious developers running agentic AI over sensitive source files and documents, Engineers who want a local coding assistant behind an OpenAI-compatible endpoint, Researchers testing open models such as Kimi K3, Muse Glimmer 30B and DeepSeek V4 Flash. Free to start; paid plans from $20/mo.
What's new in LM Studio
Checked yesterdayAcross the latest 10 updates: 4 feature updates, 3 launches, 2 changelog entries and 1 news mention.
LM Studio Bionic 1.1.8 released
Bionic 1.1.8 adds Windows on ARM CUDA detection, /v1/systemone and /v1/decisions endpoints, in-app Office and HTML previews, and llama.cpp 2.53.0 extension packs. Requires LM Studio 0.4.26+ for coexistence mode.
Bionic 1.1.7 adds Canvas export and context compaction
Bionic 1.1.7 adds Canvas export to PNG, SVG and Excalidraw, plus automatic context compaction for longer image-heavy sessions. Ships llama.cpp 2.48.0 extension packs.
Splash Engine brings fastest local Qwen3.8 on Apple Silicon
LM Studio covers running Splash by Inco AI locally, positioning Splash Engine as the fastest local Qwen3.8 inference on Apple Silicon.
Bionic adds Session References and Introspection
Bionic can search persisted session transcripts to recover details across compactions and sessions, reducing context loss in long runs.
How Auto Review works in Bionic
LM Studio details Auto Review: AST parsing, capability extraction and command matching for safer shell command execution in Bionic.
Bionic now supports skills
LM Studio added Skills to Bionic, letting users teach the agent repeatable actions.
Muse Glimmer 30B agentic model runs in Bionic
Meta's 30B agentic open-source model Muse Glimmer is available to run locally in LM Studio Bionic.
Kimi K3 in Bionic generates slides and documents
LM Studio shows Kimi K3 inside Bionic turning research into Word documents and PowerPoint decks.
DeepSeek V4 Flash available locally and hosted in US
DeepSeek V4 Flash runs in LM Studio locally and in Bionic on US-based servers with Zero Data Retention by default.
Kimi K3 available in LM Studio Bionic
MoonshotAI's Kimi K3 model is now available in LM Studio Bionic.
What people actually say about LM Studio — is it worth it?
We ran a structured research pass across product reviews, community discussions, and post-purchase forum threads to surface the patterns vendors won't publish themselves. Below: the recurring strengths, the hidden costs people mention most, and the cohort that consistently regrets adopting this tool.
105 mentions across 6 sources (Hacker News, YouTube, Product Hunt, Bluesky, Stack Overflow, Lemmy) · researched Jul 26, 2026.
Average across the 6 sources that answered — each source counts once, not each post.
- +Polished GUI makes local LLMs accessible to non-experts.
- +MLX engine with KV cache checkpointing speeds up agentic workflows.
- +MTP speculative decoding accelerates generation on supported hardware.
- +OpenAI-compatible API and SDKs simplify integration with existing tools.
- +Locally iPhone/iPad app extends local LLM use to mobile.
- −Closed-source license alienates open-source advocates.
- −Bionic update removed many advanced settings power users depend on.
- −Windows/Linux performance and stability fall behind Mac.
- −Model support limited to GGUF/MLX; fewer formats than Ollama.
- −No persistent learning; models forget corrections between sessions.
- • Pro/Enterprise pricing not publicly listed, causing uncertainty.
- • Running large models may require hardware upgrades (RAM/GPU).
Viability Score
How well maintained and how widely used is LM Studio? Built from what the vendor actually publishes (docs, changelog, tutorials, integrations, pricing), whether the site is live, and how much real users discuss it. How we calculate this
Last calculated: October 2026
How we score →Key Features
- Bionic agent creates and edits documents with automatic saving
- Coding tasks, automations and computer control via Bionic
- Real-time offline voice transcription in multiple languages
- Canvas: shared board editable by both you and Bionic (1.1.6)
- Canvas export to PNG, SVG and editable Excalidraw files (1.1.7)
- Agent introspection: reads earlier messages in its own history (1.1.4)
- Session references: search persisted transcripts and @-mention other sessions
- Automatic context compaction for long image-heavy sessions
- In-app Office and HTML previews (1.1.8)
- Windows on ARM CUDA detection (1.1.8)
- Skills: teach repeatable actions with /create-skill and /install-skill
- Auto Review parses and matches shell commands before execution
- Local LLM downloads via LM Studio Hub, run on MLX and llama.cpp
- Splash engine by Inco AI for fastest local Qwen3.8 on Apple Silicon (Mac only)
- OpenAI-compatible API (/v1/responses, /v1/chat/completions, /v1/messages, /v1/systemone, /v1/decisions)
About LM Studio
LM Studio is a desktop app for Mac, Windows and Linux that downloads open models and runs them on your own hardware through the LM Studio runtime, with MLX and llama.cpp underneath. The agent layer on top is called Bionic, and it is built for actual work rather than trivia: creating and editing documents with automatic saving on every change, coding, automations, and computer control when you ask for it. Voice input is transcribed in real time, with audio processed locally and support for multiple languages. Recent releases added real depth to the agent. Bionic 1.1.6 brought Canvas, a shared board both you and the agent edit, and 1.1.7 added Canvas export to PNG, SVG and editable Excalidraw files plus automatic context compaction for long image-heavy sessions. Introspection (1.1.4) lets Bionic read its own earlier messages, and session references let it @-mention other sessions when context has been compacted mid-project. Skills teach repeatable actions, Auto Review parses and matches shell commands before they execute, and 1.1.8 adds in-app Office and HTML previews, Windows on ARM CUDA detection and new /v1/systemone and /v1/decisions endpoints. For jobs beyond your GPU, Bionic+ sends work to US-hosted open models such as Kimi K3, GLM 5.3 and DeepSeek V4 Flash, with Zero Data Retention across LM Studio's cloud services. Meta's Muse Glimmer 30B agentic model runs locally in Bionic, and on Apple Silicon the Mac-only Splash engine from Inco AI targets the fastest local Qwen3.8 inference available. Developers get an OpenAI-compatible API, JavaScript (@lmstudio/sdk) and Python (lmstudio) SDKs, and the lms CLI. It fits developers, researchers and writers who want an agent working over sensitive files without uploading them. It is not a route to GPT- or Claude-class proprietary models, and it assumes you are willing to choose models and mind your VRAM.
Behind the Verdict
We'd reach for LM Studio when the files in question cannot leave the laptop. Contract review, patient notes, a private codebase: Bionic opens them, edits them, saves as it goes, and nothing crosses the wire unless you deliberately route the task to a cloud model. The reason it holds up past the first week is memory. Context compaction on long image-heavy sessions and introspection into its own earlier messages sound like housekeeping until you lose an hour re-explaining a project after a compaction event. Session references go further — @-mentioning another session pulls that context in. In practice that is the difference between an agent that forgets and one you can keep working with. Where it bites is model selection. You are the one choosing between a local Muse Glimmer 30B, a hosted Kimi K3, or a smaller model your GPU can actually hold. Load failures happen, VRAM math is real, and nobody at LM Studio will pick for you. Budget matters too. The Free tier is genuinely usable for local work, but Bionic+ and Pro meter cloud usage through credits, so heavy agentic runs on frontier models can outrun the headline monthly. Organizations get centralized billing for inference credits; per-seat subscription management is still on the roadmap per the vendor's own FAQ. The closest alternative for most buyers is Ollama plus a separate coding agent. Ollama is leaner and gets you a bare local endpoint fast, but you assemble the agent, the canvas and the document workflow yourself. LM Studio ships those as one app. If all you want is a local API endpoint, the leaner runtime still wins. On Apple Silicon, Splash is worth a look if Qwen3.8 speed is the bottleneck — it is Mac-only, so Windows and Linux users skip that particular advantage entirely. Who should pass: anyone whose
Researching LM Studio? Get your full AI stack in 60 seconds.
Free, no signup — tell us your goal and get tools matched to your budget & existing stack.
Real-world workflow fit
Concrete scenarios for the personas LM Studio actually fits — and what changes day-one when you adopt it.
Downloads a Qwen3.8 or Kimi model through LM Studio Hub, enables the Splash engine in Bionic 1.1.5, and asks Bionic to refactor a module in a local repo with Auto Review handling shell approvals.
Outcome: Code changes land on disk with every edit auto-saved and the source never leaves the machine.
Loads documents into Bionic, uses PDF processing for pages with large embedded images, and asks the agent to summarize and cross-reference across files using @ mentions.
Outcome: Summaries and extracted notes stay on-device, with no upload to a third-party inference provider.
Creates an organization for centralized credit billing, routes demanding tasks to US-hosted Kimi K3, GLM 5.3, or DeepSeek V4 Flash, and keeps routine work local on the Free tier.
Outcome: One bill covers cloud usage while everyday agent work stays free and local.
Use Cases
- Run local LLMs for coding assistance and data analysis with no internet connection
- Keep sensitive documents on-device while an agent edits and saves them
- Use Bionic to automate shell tasks and computer control with approval prompts
- Transcribe speech in real time offline on Mac, Windows, and Linux
- Run sessions on headless servers and expose them via the OpenAI-compatible API
- Prototype agentic workflows with KV cache checkpointing for long context
- Teach repeatable agent actions once with skills and reuse them across sessions
Models Under the Hood
as of 2026-10-16
Limitations
- Local inference is the core of the product, with cloud inference on Bionic+ ($20/mo) and Pro ($100/mo) being US-hosted open-source models under zero data retention.
- The Free tier includes the Bionic Agent, local model support, offline voice transcription, LM Link for up to 5 devices, and only limited web search.
- Team plans are not fully available: only centralized billing for inference credits via organizations exists, with subscription plans for teams coming soon.
- Web search uses ZDR, but querying and processing web content carries its own privacy and security risks, and LM Studio advises treating web content as untrusted.
as of 2026-09-29
Verification history
We have re-verified LM Studio 20 times since . Each pass re-reads the vendor's own pages and re-checks every listed field against that evidence; passes where nothing had changed are marked as such.
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
Showing the 6 most recent of 20 verification passes.
Free to cite with attribution — this page re-verifies continuously.
12-month cost
Project the real annual outlay, including the implied monthly cost when only an annual tier is published.
Vendor list price only. Add-on usage, seat overages, and contract minimums are surfaced under Hidden costs & gotchas.
Plans compared
For each published LM Studio tier: who it actually fits, and what it adds vs. the previous tier. Cross-reference the cost calculator above for projected annual outlay.
Free
$0
Ideal for
Individual developers and researchers who want Bionic Agent, local LLM inference, and offline voice transcription at no cost.
What this tier adds
Free entry point: no payment, includes Bionic Agent, llama.cpp and MLX local inference, offline voice transcription, and LM Link for up to 5 devices, with limited web search.
Bionic+
$20 USD/month
Ideal for
Individual power users who hit the limits of local hardware and want US-hosted frontier open models with zero data retention.
What this tier adds
Adds US-hosted open source models including Kimi K3, GLM 5.3, and DeepSeek V4 Flash, discounted bulk tokens, and full web search with page extraction for $20/mo.
Pro
$100 USD/month
Ideal for
Heavy agent users running long, frequent cloud-assisted sessions where the Bionic+ allowance runs out.
What this tier adds
Adds 5× usage limits over Bionic+, discounted bulk tokens, and early access to new features for $100/mo.
Cloud Credits
Pay-as-you-go
Where the pricing makes sense
The company stage and team size where LM Studio's pricing actually pencils out — and where peers do it cheaper.
LM Studio's Free tier is more capable than most local runtimes at $0: full Bionic Agent, local inference, offline voice transcription, and LM Link for five devices. Bionic+ at $20/mo targets individuals who need US-hosted frontier open models with ZDR, and Pro at $100/mo adds 5× usage limits for heavy agent users. Both cloud tiers are month-to-month with no commitment, and credits cover overage. Against cloud-only assistants that charge similar monthly rates for closed models, LM Studio's edge
Setup time & first value
How long it actually takes to get something useful out of LM Studio — broken out by persona, not the marketing-page minute.
Downloading the app and running a first local chat takes a few minutes on a machine that already has a compatible GPU or Apple Silicon. First agent task adds time to download a model — measured in minutes for small models and considerably longer for large ones. Bringing a team onto centralized cloud billing is mostly administrative through organization setup; per-seat team subscriptions are still
Switching to or from LM Studio
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
- →From Ollama: point LM Studio at the same GGUF models and use the OpenAI-compatible API endpoint to swap tooling without rewriting clients.
- →From a cloud-only assistant: move document and coding tasks to Bionic for local execution and keep frontier open models on Bionic+ for what local hardware cannot run.
- →From ChatGPT desktop workflows: use skills and @ mentions to rebuild repeatable prompts as reusable agent actions.
- →From a plain local runner: add Canvas and session introspection to keep project context across long agentic sessions.
- ↗To Ollama: export your GGUF model files and run them under Ollama if you only need model serving without the agent layer.
- ↗To a hosted cloud assistant: move to a service running GPT-4- or Claude-class models if your work depends on those specific models.
- ↗To a managed enterprise AI platform: move if you need per-seat subscription management rather than credit-based organization billing.
Integrations
Resources & Guides
Tutorials & Learning
YouTube returned 6 videos for “LM Studio”, and we withheld 5: 5 could not be judged, because “LM Studio” is a single word that other videos use for other things. Showing the 1 we can prove is about LM Studio.
Official links
Tools that pair well with LM Studio
Common stack mates teams adopt alongside LM Studio, with the specific reason each pairing earns its keep.
Ollama
Ollama runs open-weight LLMs locally or in the cloud with one command and an OpenAI-compatible API.
Mlx Serve
Free open-source local AI server that runs LLMs, image, music, video, and 3D generation on your own Apple Silicon Mac.
Atomic Chat
Atomic Chat runs 1,000+ local LLMs offline on your own hardware — free, open-source, no account, no rate limits.
Featured Head-to-Head Comparisons
Lm Studio vs Spider Cloud
These are not competitors — they are two halves of a stack, and nobody should be choosing one over the other. Pick LM Studio if your problem is where inference runs: you want open models and the Bionic agent executing on your own Mac, Windows, or Linux machine with nothing leaving your device. Pick Spider Cloud if your problem is where web data comes from: you need rendered pages, whole-site crawls, and SERP results as markdown or JSON inside your agent or RAG pipeline. A team building a private agent often ends up wanting both — Spider Cloud via its MCP server to fetch, LM Studio to reason locally — which is the clearest proof they aren't substitutes.
Lm Studio vs Polycam
These tools don't compete — they solve unrelated problems, and no buyer shortlists both. If your problem is running open models locally (GLM, Kimi, DeepSeek, Qwen3.8) for private agentic work, LM Studio is the pick: it's free on your hardware, with pay-as-you-go cloud inference when local isn't enough. If your problem is turning a room, building, prototype, or set into a 3D model or floor plan from a phone or drone, Polycam is the pick. Choose based on your job, not on a head-to-head — there isn't one.
Lm Studio vs Praktika
These two products never appear on the same shortlist. LM Studio is for a developer or privacy-focused professional who wants an agent running open models (GLM, Kimi, DeepSeek) on their own machine, with Skills, MCP, an OpenAI-compatible API, and now Linux voice transcription in Bionic 1.1.3 and the Mac-only Splash engine for Qwen3.8 in 1.1.5. Praktika is for an intermediate language learner who needs unlimited zero-stakes speaking reps with a named AI tutor, quoted around $8/month against the $30–60/hour a human tutor charges. If you're weighing them against each other, you're solving the wrong problem — pick the one that matches your actual job.
Alternatives to LM Studio
View all
Ollama
Ollama runs open-weight LLMs locally or in the cloud with one command and an OpenAI-compatible API.
Mlx Serve
Free open-source local AI server that runs LLMs, image, music, video, and 3D generation on your own Apple Silicon Mac.
Atomic Chat
Atomic Chat runs 1,000+ local LLMs offline on your own hardware — free, open-source, no account, no rate limits.
Frequently Asked Questions
Categories
Topics
Used LM Studio? Help shape our editorial sentiment research.