Local-first knowledge work
People who want one desktop agent for code and documents while keeping suitable tasks and models on their own Windows computer.
Independent tool overview
LM Studio Bionic at a glance
LM Studio Bionic is a separate Windows desktop app from the makers of LM Studio that uses local, remote, or LM Studio cloud models to work on code, documents, research, voice input, and repeatable agent skills. The app and local inference are free, while optional cloud models use prepaid credits with per-token pricing. It is unusually flexible for privacy-conscious work, but the privacy boundary depends on the selected model and tools, and its ability to run shell commands or change external files requires the same review, backups, and least-privilege controls as any coding or computer-use agent.
Visit the official LM Studio Bionic site ↗
- Product
- Separate agent app from the makers of LM Studio
- Current platform
- Windows on the public Bionic download page
- Model locations
- Local on-device, remote through LM Link, or LM Studio Secure Cloud
- Local runtime
- LM Studio's llama.cpp and MLX model runtimes
- Core work
- Coding, research, documents, presentations, spreadsheets, PDFs, automations, and computer control
- Voice
- Offline local transcription plus a cross-app voice keyboard
- Agent skills
- Reusable SKILL.md instructions compatible with the Agent Skills format
- Local price
- Free app and local inference; hardware and electricity are user costs
- Cloud price
- Prepaid credits billed per 1 million input, cached-input, and output tokens
- Cloud privacy
- Zero Data Retention and no model training claimed; cloud processing is US-based
- LM Link
- End-to-end encrypted remote model access; up to five devices on the current free plan
- Reviewed
- August 31, 2026 from current LM Studio product, docs, pricing, privacy, download, and terms pages
Overview
What LM Studio Bionic is
LM Studio Bionic is an agent-focused desktop app introduced in July 2026. It is distinct from the main LM Studio application: Bionic organizes work into projects and parallel sessions, while LM Studio remains the broader local-model runtime and developer platform.
Each Bionic session can run against a model on the same computer, a model on another device through LM Link, or an LM Studio Secure Cloud model. Local use can keep prompts and files on the device; cloud models and web search send the required request data to LM Studio or its providers for transient processing under the company's Zero Data Retention policy.
For coding, Bionic indexes a selected repository, searches and explains code, edits multiple files, runs shell tools, shows inline diffs, and can execute focused tests. For document work, it can create or edit reports, presentations, spreadsheets, PDFs, and other project files inside a managed sandbox with previews and checkpoints.
The document sandbox is not a universal containment boundary. LM Studio's documentation says files dragged in from outside a project stay in their original locations and edits are applied directly to those originals. Work on copies, maintain backups, and inspect changes before trusting rollback.
Bionic also supports local offline speech transcription and a voice keyboard that can insert text into other applications. Reusable Agent Skills use the open SKILL.md format and can be created, installed from URLs, or shared with compatible apps such as Codex and Claude Code.
The current public Bionic download page lists Windows. Do not infer macOS or Linux support from the separate LM Studio app's broader platform support. Because Bionic is a new product, availability, model catalog, cloud rates, and the forthcoming Bionic Pass may change quickly.
Use cases
Who LM Studio Bionic is best for
The strongest fit depends on the job you need the product to complete, not the size of its feature list.
Mixed model workflows
Users who want to choose between private local models, a stronger remote workstation, and pay-as-you-go frontier cloud models per session.
Repository exploration
Developers who need a conversational agent to search, explain, refactor, test, and review a local codebase.
Office deliverables
Teams creating drafts of reports, decks, spreadsheets, and PDFs with previews and checkpoint-based iteration.
Repeatable agent workflows
Power users who want portable SKILL.md instructions for recurring research, writing, review, or file-format tasks.
Remote local inference
Users with a more capable second device who want encrypted access to its models through LM Link without opening a public port.
Capabilities
Core LM Studio Bionic features
Local, remote, and cloud models
Selects where each session runs instead of forcing every task through one hosted provider.
Agentic coding
Indexes a chosen working directory, searches code, edits files, runs shell commands and tests, and exposes inline and Git diffs.
Document workspace
Creates and edits documents, presentations, spreadsheets, PDFs, and related project files with previews.
Parallel sessions
Runs multiple sessions within a project and can use sub-sessions to investigate parts of a larger task.
Checkpoints and rollback
Captures project states that can help undo agent changes, while backups and version control remain necessary.
Optional web search
Lets the agent retrieve current web information when enabled; requests leave the device and web content can carry prompt-injection risks.
Offline voice input
Transcribes speech locally and can insert dictated text at the active cursor across applications.
Reusable skills
Loads, creates, and installs standard SKILL.md packages with instructions and supporting resources.
LM Link
Connects Bionic to models running on another LM Studio device through an end-to-end encrypted mesh.
Zero-retention cloud option
Offers hosted models whose prompts and responses are processed transiently under LM Studio's stated ZDR commitments.
Process
How the LM Studio Bionic workflow works
Step 1
Choose the execution boundary
Decide whether the task can stay local, needs a remote private machine, or justifies sending data to a cloud model or web search.
Step 2
Check model and content rights
Verify the model publisher, license, allowed use, source-file rights, privacy restrictions, and any contractual or regulated-data requirements.
Step 3
Prepare an isolated workspace
Use a dedicated Git branch or worktree for code and copies of important documents; remove secrets and keep tested backups outside the agent's reach.
Step 4
Grant the minimum access
Select only the required folder, avoid production credentials and broad filesystem permissions, and keep high-risk systems disconnected.
Step 5
Define a bounded task
State the goal, allowed files, prohibited actions, acceptance checks, output format, sources, and points that require human approval.
Step 6
Inspect before changing
Ask Bionic to explain the relevant code or documents, surface assumptions, and propose a change set before enabling edits.
Step 7
Review skills and inputs
Read every installed or generated skill and treat repositories, web pages, and documents as potentially hostile instructions.
Step 8
Run a small first pass
Test the selected model on a representative low-risk task and confirm tool use, speed, context handling, and output quality.
Step 9
Inspect every change
Review inline diffs, Git diffs, external-file modifications, generated formulas, links, citations, and hidden document metadata.
Step 10
Verify with independent checks
Run focused tests, linters, security scans, calculations, accessibility checks, rendering checks, and source verification appropriate to the deliverable.
Step 11
Control cloud spend
Choose the least expensive capable model, watch input and output usage, limit long agent loops, and maintain only the credits you are prepared to consume.
Step 12
Approve and publish manually
Require an accountable person to accept code, documents, business decisions, or external actions before merging, sending, deploying, or publishing.
Cost
LM Studio Bionic pricing and free plan
The Bionic app, local models, offline voice transcription, limited signed-in web search, and LM Link for up to five devices are currently listed at $0. Cloud models use purchased credits with separate input, cached-input, and output rates. Rates below are per 1 million tokens and were checked August 31, 2026; the in-app selector and live pricing page are authoritative. A Bionic Pass is announced as coming soon but is not yet a purchasable plan.
Free
$0
For Bionic with models on the same device or connected through the current free LM Link allowance.
- Bionic agent application
- Local LLMs through llama.cpp and MLX
- Offline voice transcription
- Signed-in web search with limits and Zero Data Retention
- LM Link for up to five devices
- Local hardware, storage, electricity, and model downloads are not included costs
Budget cloud models
$0.13–$0.15 input / $0.26–$0.50 output per 1M tokens
Lower-cost usage-priced cloud models paid from credits.
- DeepSeek V4 Flash: $0.13 input, $0.028 cached input, $0.26 output
- GLM-5.3-Flash: $0.15 input, $0.03 cached input, $0.50 output
- Prices are per 1 million tokens
- US-based inference with Zero Data Retention
- Model catalog and rates can change
Standard cloud models
$0.95–$1.74 input / $3.48–$4.50 output per 1M tokens
Higher-priced cloud models for more demanding agent work.
- Kimi K2.6 and Kimi-K2.7-Code: $0.95 input, $0.16 cached input, $4.00 output
- DeepSeek V4 Pro 0813: $1.32 input, $0.132 cached input, $3.96 output
- GLM-5.3: $1.40 input, $0.14 cached input, $4.40 output
- GLM-5.2: $1.50 input, $0.30 cached input, $4.50 output
- DeepSeek V4 Pro: $1.74 input, $0.15 cached input, $3.48 output
- Prices are per 1 million tokens
Premium cloud model
$3.00 input / $15.00 output per 1M tokens
The highest listed current rate in LM Studio's cloud catalog.
- Kimi K3: $3.00 input, $0.30 cached input, $15.00 output
- Prices are per 1 million tokens
- Output-heavy agent sessions can cost substantially more than short chats
- Cloud use requires an account, billing, credits, and internet access
Bionic Pass
Coming soon
An announced future subscription whose price and included usage are not yet published.
- Do not treat this as currently available
- Confirm launch price, limits, renewal terms, and credit treatment before purchase
Pricing checked . Check current pricing at the source ↗
Assessment
LM Studio Bionic strengths and limitations
Where it stands out
- One agent interface can use local, encrypted remote, or hosted models per session
- Local inference can keep prompts, files, and model execution on the user's machine
- The core app and local-model path are free
- Coding workflow includes repository search, shell tools, tests, inline diffs, and Git awareness
- Document workflow covers several office formats with previews and checkpoints
- Offline voice transcription avoids sending audio to a hosted transcription provider
- Standard Agent Skills can transfer repeatable workflows across compatible tools
- LM Link avoids exposing a public internet port and uses end-to-end encryption
- Cloud pricing is published per model and token type instead of hidden behind an opaque allowance
- Current terms permit personal and internal business use
What to consider
- The current public Bionic download is Windows-only
- It is a new product, so feature stability, documentation, pricing, and platform availability may change quickly
- Local models require downloads and enough memory, storage, compute, and compatible hardware
- Local models vary widely in reasoning, tool use, context, hallucination rate, safety behavior, and license
- Open-model availability does not mean every model is open source or permitted for every commercial use
- Cloud models and web search send request data off-device even under a Zero Data Retention promise
- Zero Data Retention does not eliminate transit, endpoint, provider, account, output-leakage, or local-log risks
- LM Studio says cloud inference is US-based, which may not fit residency or regulated-data requirements
- A DPA, subprocessor review, retention evidence, access controls, and security review may still be required for sensitive business use
- External files added to document projects remain in place and may be edited directly despite the managed sandbox
- Shell, code-editing, automation, and computer-control capabilities can delete data, leak secrets, introduce vulnerabilities, or trigger unintended side effects
- Checkpoints may not capture every external effect and do not replace Git, backups, or independent recovery
- Installed skills, repositories, documents, and web pages can contain malicious or conflicting instructions
- Cloud agent loops, long context, and verbose output can consume credits faster than expected
- LM Link encryption does not protect against a compromised account, endpoint, remote host, model, or dependency
- Voice input can be inserted into the wrong active field and should be checked before sending
- Generated code, calculations, citations, presentations, spreadsheets, and PDFs require domain-specific human review
- The desktop software is proprietary and its terms restrict redistribution, service-bureau use, reverse engineering, and unsupported integrations
Compare
LM Studio Bionic alternatives
The right alternative depends on the specific output, workflow, controls and budget your project requires.
Coding
Ollama
Choose Ollama for a simpler local-model runtime and command-line or API foundation without Bionic's document-agent workspace.
Explore Ollama →Coding
Claude Code
Choose Claude Code for a mature terminal-centered coding agent with Anthropic's hosted models and development workflow.
Explore Claude Code →Coding
Cursor
Choose Cursor for an AI-first code editor with integrated completions, repository context, and agent features.
Explore Cursor →Business Operations
ChatGPT
Choose ChatGPT for a broad hosted assistant with mature web, mobile, voice, research, file, image, and agent capabilities.
Explore ChatGPT →Questions
LM Studio Bionic FAQs
What is LM Studio Bionic?
LM Studio Bionic is a separate Windows desktop agent from the makers of LM Studio. It works on code, research, documents, files, voice input, and repeatable skills using a local, remote, or LM Studio cloud model.
Is LM Studio Bionic the same as LM Studio?
No. Bionic is the newer agent application, while LM Studio is the established application and runtime for discovering, downloading, serving, and working with local models. The two share technology and can work together.
Is LM Studio Bionic free?
The Bionic app and local-model path are currently free. Offline voice, limited signed-in web search, and LM Link for up to five devices are also listed on the free plan. Optional cloud models consume purchased credits at published per-token rates.
Does LM Studio Bionic work offline?
Local models and offline voice transcription can work without sending prompts or audio to a cloud model once the necessary app and model files are installed. Cloud models, web search, downloads, account functions, and remote connectivity require a network.
Is LM Studio Bionic available on macOS or Linux?
As of August 31, 2026, the public Bionic download page lists Windows. The separate LM Studio app supports additional platforms, but that does not establish Bionic support on those systems.
Is LM Studio Bionic private?
It can be highly private when the selected model and all processing stay local. Cloud models and web search necessarily transmit request data for processing; LM Studio says these paths use Zero Data Retention and are not used for training, but organizations should still verify contractual, residency, security, and compliance requirements.
Can LM Studio Bionic edit my code?
Yes. After you enable coding and choose a folder, it can index the repository, search and explain code, edit files, run shell commands and tests, and show diffs. Use an isolated branch or worktree, remove secrets, review every diff, and test before merging.
Are Bionic document projects fully sandboxed?
Generated project files use a managed sandbox, but LM Studio's documentation says external files stay at their original paths and edits are applied directly to the originals. Work on copies and maintain backups.
What are Bionic Skills?
Skills are reusable instructions and optional supporting files in the Agent Skills SKILL.md format. Bionic can create, install, and load them, including compatible skills used by Codex or Claude Code. Review any skill before granting it access to important data or tools.
What does LM Studio Zero Data Retention mean?
LM Studio says prompts and responses sent through its cloud models and web search are processed transiently, are not retained by LM Studio or providers under equivalent commitments, and are not used for training. It does not mean the request never leaves the device or remove all endpoint, account, network, output, and compliance risks.
Can businesses use LM Studio Bionic?
The current desktop terms grant use for personal and internal business purposes. They restrict redistribution, service-bureau or SaaS use, reverse engineering, and integrations outside published interfaces. Businesses should review the current terms, model licenses, privacy terms, and any required enterprise agreement.
How should I control Bionic cloud costs?
Start with a local or lower-cost model, keep prompts and context bounded, monitor cached-input and output usage, stop unproductive agent loops, and buy only the credits you are prepared to spend. Recheck the in-app price before a long run.
Bottom line
Our LM Studio Bionic verdict
LM Studio Bionic is one of the more ambitious local-first desktop agents: it combines code, office-file creation, voice, portable skills, and a real choice among on-device, encrypted remote, and metered hosted models. The free local path is compelling for Windows users who can supply suitable hardware and tolerate a new product's rough edges. Its strongest privacy claim applies only when work truly stays local; cloud search and models cross that boundary, and external document edits plus shell access make disciplined isolation and review essential. Use it for supervised, reversible work—not unattended production changes or unverified high-stakes output.
Visit LM Studio Bionic website ↗Similar Coding Tools
View all Coding AI tools →
Canva Code
Canva Code 2.0 - Canva's vibe-coding tool with drag-and-drop editing
Laguna S 2.1
Laguna S 2.1 - Poolside's open-weight coding model with a 1M-token context
Grok Build
Grok Build - SpaceXAI's newly open-sourced coding agent and CLI for complex coding work
Cursor Router
Cursor Router - Cursor's auto-router picking the cheapest capable model per task
AI training for the future of work.
Get access to all our AI courses, hundreds of real-world AI use cases, live expert-led workshops, an exclusive network of AI early adopters, and more.
AI Courses
Get unlimited access to all of our current & upcoming industry-specific AI courses for the duration of your subscription.
Daily Guides
To keep up with the rapid pace of AI, our team publishes AI implementation guides daily. Our library contains 300+ practical use cases to automate real-world work.
Workshops
Join weekly, live, interactive sessions with industry leaders who are at the forefront of AI for hands-on implementation guidance and exclusive insights.
Community
Network with an exclusive community of AI-first professionals who are working smarter with AI. Learn how early adopters are using AI in their work and businesses.