LM Studio launches Bionic, agentic app for local and cloud open-source models
LM Studio released Bionic, a Mac app that runs open-source AI models locally or via cloud for coding, document processing, and research tasks. The app includes offline voice transcription using Mistral's Voxtral model and supports models like GLM 5.2 and Kimi K2.7 Code for codebase editing.
LM Studio launches Bionic, agentic app for local and cloud open-source models
LM Studio released Bionic on July 16, 2026, a Mac application that combines local model execution with cloud-based open-source models for coding, research, and document work.
Bionic extends LM Studio's existing local model runtime by adding access to larger open-source models through LM Studio Secure Cloud. According to the company, cloud requests operate under a Zero Data Retention policy, with no data storage after processing completes. Using cloud models requires an LM Studio account with billing enabled. Specific pricing was not disclosed.
Technical capabilities
For development work, Bionic creates Code projects linked to local folders, supporting models including GLM 5.2 and Kimi K2.7 Code for codebase investigation, editing, and debugging. The app includes inline diffs and agentic code search.
The app ships with Mistral AI's Voxtral model for multilingual, offline voice transcription. A voice keyboard feature works system-wide, allowing dictation wherever the cursor is positioned, with all transcription processed locally on-device.
For document work, Bionic processes PDFs, presentations, and spreadsheets in a sandboxed environment. The app includes native web search, automatic checkpoints for reviewing or rolling back changes, and in-app file previews. LM Studio stated that support for additional file type previews is coming soon.
Model compatibility
Bionic maintains LM Studio's existing Model Search tool, which matches available models to Mac hardware specifications. The app supports both local execution through LM Studio's runtime and cloud-based models for tasks requiring more computational resources, including reasoning, tool-calling, and long-context operations.
What this means
Bionic positions LM Studio as a desktop alternative to cloud-only AI agent platforms by prioritizing local execution and user control over data and costs. The combination of local and cloud model support addresses the compute limitations of consumer hardware while maintaining the privacy benefits of on-device processing. The inclusion of offline voice transcription and system-wide dictation differentiates it from browser-based tools, though adoption will depend on whether open-source models can match the capabilities of proprietary alternatives in real-world coding and document workflows.
Related Articles
Microsoft will turn Windows Search into a Copilot-connected command interface this fall
Microsoft announced a redesigned Windows Search menu that accepts short typed commands to change system settings and can converse with the new Copilot app without launching it. The company says the update arrives this fall. Model, pricing and availability details were not disclosed.
Liquid AI Publishes GGUF Builds of LFM2.5-24B-A2B, Including a 13.6 GB Quantization-Aware Q4_0 Checkpoint
Liquid AI has published GGUF builds of LFM2.5-24B-A2B, a 24B-parameter model on the lfm2moe architecture, in sizes from 13.6 GB (Q4_0) to 48 GB (BF16). The Q4_0 file is a Quantization-Aware Distillation checkpoint trained to recover accuracy lost to 4-bit quantization.
Google builds Antigravity into Gemini app for macOS, limited to AI Ultra subscribers
Google has integrated Antigravity, its agent-first development platform, directly into the Gemini app for macOS as part of the Gemini 4 Argon rollout. Access requires an AI Ultra subscription, while the new model itself requires AI Pro or above. The update also renames and reorganizes several app features, including Gemini Spark, Skills and Connectors.
Google DeepMind: Gemini 4 Argon scores 87% on ERQA embodied reasoning, targets full robotics stack
Google DeepMind says Gemini 4 Argon scores 87% on the Embodied Reasoning (ERQA) benchmark using pure reasoning, ahead of every other frontier model. A new robotics post demonstrates robot control, Real2Sim reconstruction, physical understanding, and self-improvement workflows. Pricing, context window, and availability were not disclosed.
Comments
Loading...