{"id":511,"date":"2026-03-06T13:45:50","date_gmt":"2026-03-06T05:45:50","guid":{"rendered":"https:\/\/connectword.dpdns.org\/?p=511"},"modified":"2026-03-06T13:45:50","modified_gmt":"2026-03-06T05:45:50","slug":"liquid-ai-releases-localcowork-powered-by-lfm2-24b-a2b-to-execute-privacy-first-agent-workflows-locally-via-model-context-protocol-mcp","status":"publish","type":"post","link":"https:\/\/connectword.dpdns.org\/?p=511","title":{"rendered":"Liquid AI Releases LocalCowork Powered By LFM2-24B-A2B to Execute Privacy-First Agent Workflows Locally Via Model Context Protocol (MCP)"},"content":{"rendered":"<p>Liquid AI has released <strong>LFM2-24B-A2B<\/strong>, a model optimized for local, low-latency tool dispatch, alongside <strong>LocalCowork<\/strong>, an open-source desktop agent application available in their <a href=\"https:\/\/github.com\/Liquid4All\/cookbook\/tree\/main\/examples\/localcowork\" target=\"_blank\" rel=\"noreferrer noopener\">Liquid4All GitHub Cookbook<\/a>. The release provides a deployable architecture for running enterprise workflows entirely on-device, eliminating API calls and data egress for privacy-sensitive environments.<\/p>\n<h3 class=\"wp-block-heading\"><strong>Architecture and Serving Configuration<\/strong><\/h3>\n<p>To achieve low-latency execution on consumer hardware, LFM2-24B-A2B utilizes a Sparse Mixture-of-Experts (MoE) architecture. While the model contains 24 billion parameters in total, it only activates approximately 2 billion parameters per token during inference.<\/p>\n<p>This structural design allows the model to maintain a broad knowledge base while significantly reducing the computational overhead required for each generation step. <strong>Liquid AI stress-tested the model using the following hardware and software stack:<\/strong><\/p>\n<ul class=\"wp-block-list\">\n<li><strong>Hardware:<\/strong> Apple M4 Max, 36 GB unified memory, 32 GPU cores.<\/li>\n<li><strong>Serving Engine:<\/strong> <code>llama-server<\/code> with flash attention enabled.<\/li>\n<li><strong>Quantization:<\/strong> <code>Q4_K_M GGUF<\/code> format.<\/li>\n<li><strong>Memory Footprint:<\/strong> ~14.5 GB of RAM.<\/li>\n<li><strong>Hyperparameters:<\/strong> Temperature set to 0.1, top_p to 0.1, and max_tokens to 512 (optimized for deterministic, strict outputs).<\/li>\n<\/ul>\n<h3 class=\"wp-block-heading\"><strong>LocalCowork Tool Integration<\/strong><\/h3>\n<p>LocalCowork is a completely offline desktop AI agent that utilizes the Model Context Protocol (MCP) to execute pre-built tools without relying on cloud APIs or compromising data privacy, logging every action to a local audit trail. The system includes 75 tools across 14 MCP servers capable of handling tasks like filesystem operations, OCR, and security scanning. However, the provided demo focuses on a highly reliable, curated subset of 20 tools across 6 servers, each rigorously tested to achieve over 80% single-step accuracy and verified multi-step chain participation.<\/p>\n<figure class=\"wp-block-embed is-type-video is-provider-youtube wp-block-embed-youtube wp-embed-aspect-4-3 wp-has-aspect-ratio\">\n<div class=\"wp-block-embed__wrapper\">\n<\/div>\n<\/figure>\n<p>LocalCowork acts as the practical implementation of this model. It operates completely offline and comes <strong>pre-configured with a suite of enterprise-grade tools:<\/strong><\/p>\n<ul class=\"wp-block-list\">\n<li><strong>File Operations:<\/strong> Listing, reading, and searching across the host filesystem.<\/li>\n<li><strong>Security Scanning:<\/strong> Identifying leaked API keys and personal identifiable information (PII) within local directories.<\/li>\n<li><strong>Document Processing:<\/strong> Executing Optical Character Recognition (OCR), parsing text, diffing contracts, and generating PDFs.<\/li>\n<li><strong>Audit Logging:<\/strong> Recording every tool call locally for compliance tracking.<\/li>\n<\/ul>\n<h3 class=\"wp-block-heading\"><strong>Performance Benchmarks<\/strong><\/h3>\n<p>Liquid AI team evaluated the model against a workload of 100 single-step tool selection prompts and 50 multi-step chains (requiring 3 to 6 discrete tool executions, such as searching a folder, running OCR, parsing data, deduplicating, and exporting).<\/p>\n<h4 class=\"wp-block-heading\"><strong>Latency<\/strong><\/h4>\n<p>The model averaged <strong>~385 ms per tool-selection response<\/strong>. This sub-second dispatch time is highly suitable for interactive, human-in-the-loop applications where immediate feedback is necessary.<\/p>\n<h4 class=\"wp-block-heading\"><strong>Accuracy<\/strong><\/h4>\n<ul class=\"wp-block-list\">\n<li><strong>Single-Step Executions:<\/strong> 80% accuracy.<\/li>\n<li><strong>Multi-Step Chains:<\/strong> 26% end-to-end completion rate.<\/li>\n<\/ul>\n<h3 class=\"wp-block-heading\"><strong>Key Takeaways<\/strong><\/h3>\n<ul class=\"wp-block-list\">\n<li><strong>Privacy-First Local Execution:<\/strong> LocalCowork operates entirely on-device without cloud API dependencies or data egress, making it highly suitable for regulated enterprise environments requiring strict data privacy.<\/li>\n<li><strong>Efficient MoE Architecture:<\/strong> LFM2-24B-A2B utilizes a Sparse Mixture-of-Experts (MoE) design, activating only ~2 billion of its 24 billion parameters per token, allowing it to fit comfortably within a ~14.5 GB RAM footprint using <code>Q4_K_M GGUF<\/code> quantization.<\/li>\n<li><strong>Sub-Second Latency on Consumer Hardware:<\/strong> When benchmarked on an Apple M4 Max laptop, the model achieves an average latency of ~385 ms for tool-selection dispatch, enabling highly interactive, real-time workflows.<\/li>\n<li><strong>Standardized MCP Tool Integration:<\/strong> The agent leverages the Model Context Protocol (MCP) to seamlessly connect with local tools\u2014including filesystem operations, OCR, and security scanning\u2014while automatically logging all actions to a local audit trail.<\/li>\n<li><strong>Strong Single-Step Accuracy with Multi-Step Limits:<\/strong> The model achieves 80% accuracy on single-step tool execution but drops to a 26% success rate on multi-step chains due to \u2018sibling confusion\u2019 (selecting a similar but incorrect tool), indicating it currently functions best in a guided, human-in-the-loop loop rather than as a fully autonomous agent.<\/li>\n<\/ul>\n<hr class=\"wp-block-separator has-alpha-channel-opacity\" \/>\n<p>Check out the\u00a0<strong><a href=\"https:\/\/github.com\/Liquid4All\/cookbook\/tree\/main\/examples\/localcowork\" target=\"_blank\" rel=\"noreferrer noopener\">Repo<\/a> <\/strong>and<strong> <a href=\"https:\/\/www.liquid.ai\/blog\/no-cloud-tool-calling-agents-consumer-hardware-lfm2-24b-a2b\" target=\"_blank\" rel=\"noreferrer noopener\">Technical details<\/a>.\u00a0<\/strong>Also,\u00a0feel free to follow us on\u00a0<strong><a href=\"https:\/\/x.com\/intent\/follow?screen_name=marktechpost\" target=\"_blank\" rel=\"noreferrer noopener\"><mark>Twitter<\/mark><\/a><\/strong>\u00a0and don\u2019t forget to join our\u00a0<strong><a href=\"https:\/\/www.reddit.com\/r\/machinelearningnews\/\" target=\"_blank\" rel=\"noreferrer noopener\">120k+ ML SubReddit<\/a><\/strong>\u00a0and Subscribe to\u00a0<strong><a href=\"https:\/\/www.aidevsignals.com\/\" target=\"_blank\" rel=\"noreferrer noopener\">our Newsletter<\/a><\/strong>. Wait! are you on telegram?\u00a0<strong><a href=\"https:\/\/t.me\/machinelearningresearchnews\" target=\"_blank\" rel=\"noreferrer noopener\">now you can join us on telegram as well.<\/a><\/strong><\/p>\n<p>The post <a href=\"https:\/\/www.marktechpost.com\/2026\/03\/05\/liquid-ai-releases-localcowork-powered-by-lfm2-24b-a2b-to-execute-privacy-first-agent-workflows-locally-via-model-context-protocol-mcp\/\">Liquid AI Releases LocalCowork Powered By LFM2-24B-A2B to Execute Privacy-First Agent Workflows Locally Via Model Context Protocol (MCP)<\/a> appeared first on <a href=\"https:\/\/www.marktechpost.com\/\">MarkTechPost<\/a>.<\/p>","protected":false},"excerpt":{"rendered":"<p>Liquid AI has released LFM2-24&hellip;<\/p>\n","protected":false},"author":1,"featured_media":512,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[],"class_list":["post-511","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-uncategorized"],"_links":{"self":[{"href":"https:\/\/connectword.dpdns.org\/index.php?rest_route=\/wp\/v2\/posts\/511","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/connectword.dpdns.org\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/connectword.dpdns.org\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/connectword.dpdns.org\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/connectword.dpdns.org\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=511"}],"version-history":[{"count":0,"href":"https:\/\/connectword.dpdns.org\/index.php?rest_route=\/wp\/v2\/posts\/511\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/connectword.dpdns.org\/index.php?rest_route=\/wp\/v2\/media\/512"}],"wp:attachment":[{"href":"https:\/\/connectword.dpdns.org\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=511"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/connectword.dpdns.org\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=511"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/connectword.dpdns.org\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=511"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}