Full-Time
Recruiting Staff
Technical Recruiter
Liquid AI
51-200 employees
Provides generative, edge-capable foundation models
No salary listed
San Francisco, CA, USA + 1 more
More locations: United States
Hybrid
At least four days per week in the San Francisco office.
 People & HR (1) |
|---|
Get referred to Liquid AI
See people who can refer or advise you
- Proven depth in full-cycle recruiting for research, engineering, or other deeply technical roles, with the seniority to own complex searches end-to-end.
- A track record of winning competitive closes, including competing offers, equity education, total-compensation narratives, creative engagement, and in-person closes.
- Experience defining roles, rubrics, and evaluation processes from scratch and coaching hiring managers, including first-time hiring managers.
- Strong applicant-tracking-system discipline using Ashby or a comparable system, plus recruiting-data rigor involving structured screens, scorecards, calibrated loops, and accurate, reportable metrics such as time-to-fill, pass-through, and quality of hire.
- Based in San Francisco and able to work from the office at least four days per week and host candidate on-sites.
- Own full-cycle hiring across engineering, research, product, solutions architecture, go-to-market, and general and administrative roles, serving as the in-person anchor in San Francisco, running on-sites, and closing senior and executive hires face-to-face.
- Coach stakeholders and partner with senior leaders and the C-suite to define brand-new roles before searches begin.
- Identify the best talent by building evidence for why a candidate is the best fit, beyond LinkedIn availability.
- Build rigorous, role-specific evaluation processes from scratch by defining success criteria and assessing attributes and values that predict success.
- Win and close priority candidates by sharpening the value proposition, using data to unblock stalled searches, and gathering market intelligence.
- Measure quality of hire and run look-backs, maintain reportable Ashby data, document processes in Notion, and use artificial-intelligence-enabled workflows.
- Experience recruiting for artificial intelligence, machine learning, or deep-technology roles.
- Early-stage team-building experience from seed through Series C, helping build the recruiting motion rather than executing a mature playbook.
- Experience building quality-of-hire and retention loops, including onboarding check-ins, hire look-backs, and calibration back into evaluation.
- Range across product, go-to-market, general and administrative, and customer-facing technical roles, or clear evidence of the ability to flex into them.
Liquid AI builds and deploys foundation models based on liquid neural networks, focusing on efficient, on-device AI. It develops Liquid Foundation Models (LFMs), a family of generative AI models designed to be smaller and more computation-efficient than typical large language models, enabling deployment on edge devices with lower latency, better privacy, and reduced infrastructure costs. The approach includes end-to-end AI expertise and customizable architectures for enterprises that require real-time performance and private processing. Compared to traditional AI providers, Liquid AI emphasizes edge-ready, adaptable models that run efficiently on constrained hardware, and it targets enterprise-grade, private, reliable AI solutions. The company’s goal is to enable real-time, on-device AI at scale for businesses by offering compact, capable foundation models and the tools to customize them for specific applications.
Company Size
51-200
Company Stage
Series A
Total Funding
$287.5M
Headquarters
Brookline, Massachusetts
Founded
2023
Get referred to Liquid AI
See people who can refer or advise you
Simplify's Take
What believers are saying
- August 20, 2026 DSpark claims 3.18x GPU and 2.87x on-device decoding speedups.
- LFM2.5-VL-3B on August 12, 2026 improved ScreenSpot-v2 to 80.7 and RefCOCO to 87.9.
- MacPaw's August 2026 partnership and Setapp access create immediate enterprise and developer channels.
What critics are saying
- Apple can bundle stronger native on-device AI, compressing Liquid AI's differentiation by 2027.
- Open-weight rivals like Qwen and Gemma can match benchmarks, eroding pricing power fast.
- No public revenue disclosure or moat beyond model releases makes acquisition dependence an existential risk.
What makes Liquid AI unique
- MIT spinout Liquid AI ships open-weight, on-device foundation models with 34T-token training.
- August 2026 releases span text, vision, encoders, and speculative decoding for edge deployment.
- MacPaw partnership embeds Liquid models into Eney and Setapp, expanding distribution beyond APIs.
Help us improve and share your feedback! Did you find this helpful?
Benefits
Remote Work Options
Flexible Work Hours
Growth & Insights and Company News
6 month growth
↑ 2%1 year growth
↓ -2%2 year growth
↑ 5%Block launches Berd workspace as research pivot targets high stakes industrial applications. Block released Berd, an Apache 2.0 licensed agent workspace designed to orchestrate multiple models while storing data locally. This move signals a shift toward privacy-centric enterprise infrastructure that treats base models as interchangeable commodities. By open-sourcing the orchestration... Executive summary. Block released Berd, an Apache 2.0 licensed agent workspace designed to orchestrate multiple models while storing data locally. This move signals a shift toward privacy-centric enterprise infrastructure that treats base models as interchangeable commodities. By open-sourcing the orchestration layer, Block is positioning itself to define how businesses manage agentic workflows without being locked into a single provider. Why now Enterprises are increasingly skeptical of sending proprietary data to centralized labs for every inference call. This week's research reflects a broader push toward vertical reliability and efficiency, seen in new frameworks for flight safety analysis and tabular regression. There's a clear trend toward moving model capabilities out of general-purpose chatbots and into specialized, high-stakes industrial applications. What's new Block's Berd workspace allows users to swap between different models while maintaining a unified local conversation history (VentureBeat). Liquid AI released LFM 2.5 checkpoints utilizing quantization-aware distillation, targeting lower inference costs for edge deployment (Hugging Face). Researchers introduced "Chain-of-Experience," a method for continuous model improvement that moves beyond static retraining cycles (arXiv). New studies in agentic receptivity for online dating and flight safety suggest labs are testing model autonomy in complex social and physical environments (arXiv). What to watch Adoption rates of open-source agent orchestrators like Berd. If these become the enterprise standard, the "moats" around proprietary model ecosystems will continue to erode. Results from specialized LLM applications in safety-critical sectors. Success in aviation or medicine will be the leading indicator for the next wave of enterprise spending. Advancements in distillation techniques. Liquid AI's focus on smaller, efficient models is the blueprint for firms looking to escape the high margins of massive compute providers. Drafted and published autonomously by the McGauley Labs agent pipeline. No per-briefing human approval. Governed by its public style guide. Byline: McGauley Labs Drafting Model: Gemini 3.0 Pro Sources: - https://venturebeat.com/orchestration/blocks-new-apache-2-0-agent-workspace-berd-works-across-models-and-harnesses-stores-conversation-history-locally - https://huggingface.co/blog/LiquidAI/qad - https://arxiv.org/abs/2608.18027v1 - https://arxiv.org/abs/2608.18017v1 - https://arxiv.org/abs/2608.18058v1 Continue Reading: Product Launches | By: McGauley Labs Drafting model: Gemini 3.0 Pro Block launched Berd, an Apache 2.0 agent workspace that works across different models and stores conversation history locally. This release signals a shift toward developer sovereignty, moving away from the cloud-dependency that characterizes most current agent orchestration. By offering a model-agnostic layer, Block allows users to swap backends while maintaining data residency on their own hardware. The move comes as the industry pivots from experimental chatbots to persistent agentic systems that require reliable memory. Block is positioning itself as a provider of the essential infrastructure for developers who won't risk sending sensitive logs to third-party providers. This coincides with a heavy week for research, where technical refinements like the WEASEL 2.0 adaptive ensemble-size rule are attempting to make classification more stable and reproducible (per an arXiv paper). What's new Berd operates as a model-agnostic layer, allowing developers to switch between labs like OpenAI and Anthropic without rewriting integration logic (VentureBeat reported). Local storage of conversation history addresses data sovereignty hurdles that often block AI adoption in regulated industries. Researchers updated WEASEL 2.0 to include an adaptive ensemble-size rule to fix sensitivity and reproduction issues in time-series models (per arXiv 2608.18021v1). What to watch Adoption of Berd within Block's own ecosystem, specifically its projects focused on decentralized identity and payments. Whether the "local-first" trend forces cloud-native orchestration platforms to offer similar on-premise history management to remain competitive. Commercial implementation of the WEASEL 2.0 refinements in industrial time-series monitoring. Research & Development | Research labs are shifting focus toward high-stakes industrial applications and architectural efficiency to justify massive compute expenditures. This week's research highlights a push into aviation safety, structured tabular data, and the psychological barriers of agentic systems. A new arXiv paper proposes a prior-guided semantic approach to help models explain flight safety events. While generic systems struggle with the precision required for aviation, this method uses structured domain knowledge to ground model outputs. Investors should view this as a necessary step toward the utility of models in regulated industries where hallucination carries literal life-or-death consequences. Tabular data remains the primary asset for most enterprises, yet deep learning often fails to beat simple gradient-boosted trees. TabNSM introduces a Neural Sparse Mixer for tabular regression to bridge this gap. If neural architectures can finally dominate structured data, the value of centralized model training for business intelligence increases significantly compared to current fragmented methods. The delegation of personal decisions to agents remains a significant psychological hurdle. Research into agentic recommender systems in the online dating market reveals a "delegation asymmetry" where users resist offloading social interactions. This suggests that agentic adoption will likely stall at the discovery phase until labs can solve the trust deficit inherent in automated communication. Efficiency gains in spatial and temporal learning are surfacing through Memory Tree structures and Chain-of-Experience (CoE) frameworks. The former optimizes 3D question answering by querying key frames more effectively, while CoE targets the expensive cycle of model retraining. These optimizations indicate that the next phase of competition will focus on inference cost and continuous learning rather than raw parameter count. What to watch. Adoption of sparse mixers like TabNSM in cloud-native business intelligence tools from providers like Snowflake or Databricks. Regulatory feedback from the FAA or EASA regarding LLM-based reporting as a valid safety audit tool. Benchmarks for "experience-based" training as a proxy for reduced long-term compute requirements in production environments. Drafted and published autonomously by the McGauley Labs agent pipeline. No per-briefing human approval. Governed by its public style guide. Byline: McGauley Labs via Gemini 3.0 Pro Sources gathered by its internal agentic system. Article processed and written by Gemini 3.0 Pro (gemini-3-flash-preview). This digest is generated from multiple news sources and research publications. Always verify information and consult financial advisors before making investment decisions.
Liquid AI Releases LFM2.5-VL-3B: A 3B Vision-Language Model That Reads Screens, grounds objects, and calls Tools on-device. Yesterday, Liquid AI released LFM2.5-VL-3B. It is a 3.1B-parameter vision-language model built for on-device deployment. The model reads digital screens across mobile, web, and... Source and context MarkTechPost · Observe 1-12 months Aug 13, 2026, 3:56 PM Today's signal Fast orientation Trend Confidence Medium · 1-12 months Reality status Live or rolling out Release phase. This is being reported as a release, rollout, or product move rather than a hypothetical plan. The main uncertainty is adoption and consequence, not whether the move exists. Signal panel Scan the signal before you read the analysis. * Signal level - Trend * Signal strength - Medium * Time horizon - 1-12 months * Human impact - Low * Economic impact - Low * Governance impact - Low * Confidence - Medium Original signal What the source is actually reporting. What happened Yesterday, Liquid AI released LFM2.5-VL-3B. It is a 3.1B-parameter vision-language model built for on-device deployment. The model reads digital screens across mobile,... Who is involved The clearest named actors are Liquid AI Releases LFM2.5-VL-3B and Vision-Language Model That Reads Screens. The likely spillover reaches labs, institutions, and publics exposed to a larger directional shift. What changed A new model, product, feature, or capability is moving into practical circulation. It is being reported now because a new capability has moved from planning into visible release or rollout. Chip rewritten report A fuller reader version of the report. Reader version MarkTechPost reports this core fact: Yesterday, Liquid AI released LFM2.5-VL-3B. It is a 3.1B-parameter vision-language model built for on-device deployment. The model reads digital screens across... The clearest named actors are Liquid AI Releases LFM2.5-VL-3B and Vision-Language Model That Reads Screens. The likely spillover reaches labs, institutions, and publics exposed to a larger directional shift. A new model, product, feature, or capability is moving into practical circulation. It is being reported now because a new capability has moved from planning into visible release or rollout. For readers, this belongs in the AI Tools lane and the AI Models topic, which means the important details are not only who announced what, but which expectations, costs, rules, or capabilities may now move around it. The useful reading is simple: A new AI capability is moving from announcement into practical circulation. Chip interpretation What it means The reported move is simple: Yesterday, Liquid AI released LFM2.5-VL-3B. It is a 3.1B-parameter vision-language model built for on-device deployment. The model reads digital screens across mobile, web, and... Read this through The practical question is whether this becomes a repeated pattern that operators, governments, or ordinary users will need to treat as normal. Decision test Read this as a directional signal about the broader AI trajectory, not just as a short-term product update. For anyone affected by models, the useful test is whether this changes trust, cost, rules, capability, or expected human judgment after the first attention wave passes. Why this matters The consequence is more important than the headline. These are the practical consequence areas to watch if this signal repeats beyond a single article. Impact card Business impact. The business effect is limited for now. Treat this more as directional context than as an immediate budget move. Impact card Human impact. Direct human impact looks limited right now. Even so, it helps explain the direction AI systems are moving toward. Impact card AI ecosystem impact. At ecosystem level, this is a pattern signal more than a final verdict. Repeated moves of this kind are what reset the baseline over time. Who gains / who is pressured Follow the incentives, not the announcement. * Institutions that prepare early: They benefit when they build frameworks before capability pressure becomes urgent. * Long-horizon builders: They gain from understanding direction before it hardens into infrastructure or law. Who is pressured * Reactive organizations: They are exposed when they only respond after the larger system has already shifted. * Low-trust information environments: They become more fragile when capability rises without matching clarity or governance. Multiple perspectives Trust improves when the angles are visible. Builder view The key issue is whether capability is growing inside structures strong enough to keep orientation, consent, and return. Government view The concern is whether institutions can keep pace before strategic capability becomes irreversible infrastructure. Citizen view The practical question is whether ordinary people gain more agency from the shift or become more dependent on systems they cannot inspect. What humans should do Primary action: Observe. * Do not overreact to a single article. Watch for pattern repetition across other sources and follow-on moves. * Note whether this changes expectations in your lane even if it does not require action yet. * Use it as orientation, not as a reason to make rushed operational changes. Signal memory Original source Source and evidence still matter. This page is a Chip interpretation of the original article. It is not the original article. Please read the original source for the full report. Curation note: this brief uses the source link, attribution, and original Age for AI commentary. It is not permission to repost the publisher's full text, images, or reporting elsewhere. What readers are saying. No comments yet
MacPaw partners with Liquid AI to build on-device AI infrastructure for Mac apps. Table of content. The companies will combine Liquid AI's foundation models with MacPaw's local inference and memory technologies, beginning with the Eney assistant and potentially expanding the system to Setapp developers. Published: August 6, 2026 MacPaw has entered a long-term partnership with Liquid AI to develop an artificial intelligence technology stack that can run directly on Mac computers. The collaboration will combine Liquid AI's efficient foundation models with two technologies developed by MacPaw: Elix, an on-device inference system, and Mnemos, a local memory layer. MacPaw's Eney assistant will be the first product to use the combined system. The companies expect to demonstrate initial results during 2026. After deploying the technology in Eney, MacPaw plans to extend the infrastructure across its product ecosystem and potentially make it available to developers distributing software through Setapp. Partnership fact box. Eney will be the first product to use the technology. MacPaw is initially applying the new AI stack to Eney, its artificial intelligence assistant for macOS. Eney is designed to allow users to complete tasks across Mac applications through conversational instructions. MacPaw introduced the assistant in 2025 as part of its effort to create a more unified interface for software and workflows. The company wants Eney to process more tasks directly on the user's computer instead of sending every request to an external cloud server. Liquid AI will develop and adapt its foundation models for tasks performed by the assistant. These models will operate through MacPaw's Elix inference framework on Macs powered by Apple silicon. Cloud-based models will remain available when MacPaw determines that remote processing is more appropriate for a particular task. The planned AI stack has three main layers. The partnership combines three technical components that perform different functions. How the on-device system is designed to work. The planned architecture is designed to prioritize local processing while retaining access to cloud-based AI systems. Local execution can allow supported tasks to continue without an internet connection. It can also reduce the amount of personal information that must leave the computer during processing. The companies have not stated that every Eney function will operate offline. Some requests may continue to use cloud infrastructure when a remote model offers capabilities that are not available locally. Elix will handle AI inference on Apple silicon. Elix is MacPaw's framework for running artificial intelligence models directly on a device. Inference is the stage at which a trained AI model receives an input and produces an output. In Eney's case, an input could be a user request to search for information, modify a file or execute a supported action through another application. Liquid AI will optimize its models for the hardware and tasks involved rather than relying on one general configuration for every device. The models will then run through Elix on Apple silicon, the chip architecture used across Apple's current Mac product range. The partnership is focused on building an integrated system rather than simply connecting Eney to an external AI provider through a cloud application programming interface. Mnemos will provide long-term local memory. The second major MacPaw technology involved in the project is Mnemos. Mnemos is intended to give Eney persistent memory across user interactions. This could allow the assistant to retain approved contextual information instead of treating every conversation as an entirely separate session. The memory layer is being developed to operate locally. MacPaw says this approach is intended to keep personal context stored on the user's computer. Mnemos development objectives. The companies have not disclosed how much information Mnemos will store, how long individual records will be retained or which controls users will receive for reviewing and deleting saved information. Liquid AI will adapt its models for mac-based tasks. Liquid AI develops foundation models intended to operate efficiently across local devices and other computing environments. The company was founded by researchers connected to the Massachusetts Institute of Technology. Its technology is designed to reduce the computing resources required to run capable AI models. For the MacPaw partnership, Liquid AI will train and fine-tune models for the tasks Eney is expected to perform. This process may involve selecting model architectures based on the available hardware and optimizing them for specific assistant functions. Liquid AI has said that its approach begins with choosing an architecture suited to the target device before training the model. The company argues that this allows the final system to use computing resources more efficiently. Local and cloud processing will operate together. MacPaw is not planning to remove cloud-based models completely. The company intends to use a hybrid architecture in which the system selects between local and remote processing depending on the request. The companies have not published the exact criteria that will determine when a request is processed locally and when it is transferred to a cloud model. TechCrunch reported that MacPaw also wants to provide access to cloud models from companies such as Google. This could allow developers to select between MacPaw and Liquid AI's local infrastructure and external AI services. Technology deployment will begin with Eney. The companies have described Eney as the first stage of a broader deployment plan. * The first stage will involve integrating the models, inference framework and memory layer into Eney. * The second stage could expand the infrastructure to additional MacPaw products. * The third stage could make parts of the technology available to independent developers through Setapp. MacPaw has not announced a fixed release date for the developer tools. Setapp could become a distribution platform for AI apps. Setapp is MacPaw's software marketplace and subscription service for Mac applications. The platform has more than 150,000 paying users. It provides access to a collection of Mac, iOS and web applications through subscription and individual app purchasing options. MacPaw plans to place a greater focus on artificial intelligence applications within Setapp. Once the local processing architecture is finalized, the company wants to allow developers to use the on-device inference system in their own applications. This could give participating developers access to a shared AI infrastructure without requiring each developer to independently build a complete local model deployment system. Potential Setapp developer infrastructure. MacPaw is testing credit-based AI pricing. MacPaw is experimenting with a credit-based pricing model for AI operations within Setapp. Under the proposed system, users would receive or purchase credits that could be spent when performing AI-powered tasks. The number of credits used could vary based on the complexity of the requested operation. A basic AI action could require fewer credits than a task involving more extensive processing or the use of an external cloud model. MacPaw has not published credit prices, usage limits or the number of credits that would be included with existing Setapp subscriptions. Apple already provides local AI models to developers. MacPaw and Liquid AI are entering an ecosystem in which Apple already provides developers with access to on-device artificial intelligence technologies. Apple's developer tools allow supported applications to use models and machine-learning frameworks that operate on Apple hardware. MacPaw and Liquid AI are attempting to differentiate their system through models optimized for different capabilities, a customizable model layer, persistent local memory and integration with Eney and Setapp. The companies have not published direct performance comparisons between their planned system and Apple's local AI technologies. No benchmark data has been released showing differences in speed, accuracy, memory use or energy consumption. The partnership is a development project, not a finished product. MacPaw and Liquid AI have announced their technical direction, but several important details remain under development. The companies have not released final performance measurements for Eney's new local models. They have also not disclosed the model sizes, minimum Mac specifications, memory requirements or storage requirements for the finished system. Confirmed and undisclosed information. Initial results are expected during 2026. MacPaw and Liquid AI expect the first results from the partnership during 2026. The initial milestone will be the implementation of the on-device stack in Eney. A later expansion could bring the same infrastructure to other MacPaw products and independent applications distributed through Setapp. The companies have not announced when the technology will become broadly available to users or developers. Until technical specifications and performance results are released, the partnership remains an active infrastructure project rather than a completed on-device AI platform. Post comment. Be the first to post a comment!
MacPaw partners with Liquid AI to bring on-device inference to app store developers. Inside this article. MacPaw has entered a long-term partnership with Liquid AI to build a local artificial intelligence stack for macOS, beginning with the Ukrainian software company's Eney assistant and potentially expanding to developers distributing apps through Setapp. Announced on August 5, the collaboration will combine Liquid AI's compact foundation models with MacPaw's own inference and memory technologies. The goal is to move more AI processing from remote data centers onto Apple silicon, allowing supported features to respond faster, work without a constant internet connection and keep more personal information on the user's Mac. MacPaw said the first results are expected in Eney later in 2026. The companies are also designing the underlying components as shared infrastructure that could eventually be offered to thousands of Mac developers through Setapp. No public date has been set for that wider developer rollout. Eney becomes the test bed. Eney, introduced by MacPaw as a proactive assistant for macOS, will serve as the first production environment for the joint work. Rather than relying entirely on a cloud model for every request, the planned system will decide which tasks can be handled locally and which still require more powerful remote models. That hybrid design matters because desktop assistants may need access to files, application activity, preferences and conversation history. Sending all of that context to an external server creates additional privacy, latency and connectivity concerns. Local processing can reduce those dependencies, although smaller on-device models remain constrained by available memory and compute power. MacPaw founder and CEO Oleksandr Kosovan described the company's direction in clear terms: "We believe intelligence should live where people work: private by design, fast by default." Cloud systems will not disappear from Eney. MacPaw says heavier tasks can still be routed remotely when a cloud model is more suitable. The company is building a split architecture rather than presenting local inference as a complete replacement for hosted AI. Elix handles local inference. The inference layer is called Elix. MacPaw describes it as a Mac-native engine built on MLX, Apple's machine-learning framework for Apple silicon. It is being tuned with techniques such as speculative decoding, quantization and key-value cache optimization to improve model speed and efficiency on Mac hardware. Liquid AI will develop and fine-tune Liquid Foundation Models for the tasks Eney is expected to perform. The startup designs models around the hardware and operating conditions in which they will run, rather than treating deployment as a final optimization step. Liquid AI co-founder and CEO Ramin Hasani said the partnership would bring "efficient, private, on-device LFMs to millions of Mac users." Its broader edge strategy is visible in LEAP, a developer platform created to deploy models on phones, laptops, vehicles and other local hardware. Liquid AI has said LEAP can integrate supported foundation models into mobile applications with a small amount of code, reflecting its effort to make local deployment accessible beyond specialist inference teams. Mnemos adds persistent memory. The second MacPaw component, Mnemos, is a local-first memory layer intended to retain useful context across interactions. Instead of treating every request as an isolated prompt, it can help an assistant remember relevant information and use it in later tasks. Persistent memory is central to MacPaw's vision for Eney because an assistant operating across applications must understand more than the current chat. It may need to connect a file opened earlier, a recurring workflow and a preference established days before. MacPaw says Mnemos is being designed as an on-device knowledge base, reducing the need to keep that context in a hosted service. MacPaw has not yet detailed the user controls, retention settings or developer-facing permissions that will govern the final system. Setapp could become an AI platform. The partnership extends beyond a single assistant. MacPaw plans to make the models, inference framework and memory layer reusable across its product ecosystem. The longer-term opportunity is to expose those building blocks to independent Mac developers through Setapp. Setapp currently offers more than 250 apps and has over 150,000 paying users. MacPaw has been repositioning the subscription service from an app catalog into a broader development and distribution platform for AI software. Its existing AI Gateway gives developers a single API for models from providers including OpenAI, Anthropic, Google and xAI. The gateway handles provider routing, authentication and Setapp's credit system, reducing the need for each app maker to manage separate integrations and user API keys. Adding local inference would create another route. Developers could run suitable tasks directly on a Mac, call a cloud model for more demanding work, or combine both within the same application. That could support document classification, summarization, transcription, code assistance and other features involving private data. MacPaw is also experimenting with credit-based AI plans in Setapp, where usage depends on the number and complexity of operations. How local processing will be priced, metered or shared with developers has not been disclosed. A different route from cloud-only AI. The announcement arrives as software companies look for ways to control the cost and privacy implications of AI features. Cloud models remain stronger for many complex jobs, but every remote request can add server expense, network delay and data-handling obligations. On-device models change that calculation. Once downloaded, they can perform repeated tasks without paying an external provider for every inference and continue operating without connectivity. The trade-off is that developers must account for hardware differences, model size, updates and performance limits across supported Macs. Apple already provides developers with access to on-device intelligence through its own frameworks. MacPaw is positioning Elix and Liquid AI's models as a complementary stack with Mac-specific optimization, custom adapters, persistent memory and a connection to Setapp's cloud gateway. The rollout questions. For now, the partnership is an infrastructure commitment rather than a finished developer product. Eney is scheduled to show the first results later this year, while availability through Setapp remains a future step. MacPaw has not announced supported Mac models, minimum memory requirements, SDK pricing, developer revenue terms or a firm release schedule. Those details will determine whether the system becomes a practical foundation for mainstream Mac apps or remains limited to selected partners and MacPaw's own products. Still, the direction is clear: MacPaw wants Setapp to offer developers both local and cloud AI through one commercial ecosystem. If the companies deliver that architecture, app makers could gain a simpler way to build private, offline-capable features without assembling a separate inference engine, memory system, model pipeline and billing layer. The immediate test will be Eney. The larger test will be whether MacPaw can turn that internal stack into infrastructure other Mac developers can reliably ship.
MacPaw and Liquid AI Partner to Build On-Device AI for the Mac. MacPaw and Liquid AI announced a long-term partnership today to co-develop an on-device AI technology stack built specifically for the Mac. Eney, MacPaw's AI assistant for macOS, will be the first product to use the new platform. MacPaw plans to expand it to additional products and, eventually, make it available to developers through Setapp. The new AI platform combines Liquid AI's Liquid Foundation Models (LFMs), which will be designed specifically for macOS AI assistance, with MacPaw's Elix inference engine and Mnemos memory layer. Together, the technologies are designed to keep personal data on the Mac, deliver fast responses, and handle core AI tasks even without an internet connection. Cloud models will remain available where they are the better tool. "After almost two decades of building software for the Mac, MacPaw is evolving its standalone products into a connected ecosystem, with AI as a core technology we build and own," said Oleksandr Kosovan, CEO and founder of MacPaw. "We believe intelligence should live where people work: private by design, fast by default, and be able to reach the cloud when that's the better tool." Liquid AI says its foundation models are built to bring efficient AI directly to users' devices. "We build efficient foundation models and the tools around them so that companies can bring intelligence onto the devices their customers already use," said Ramin Hasani, co-founder and CEO of Liquid AI. "This partnership will bring efficient, private, on-device LFMs to millions of Mac users." The company's models are already used by Mercedes-Benz, Insilico Medicine, and Shopify for on-device AI workloads. MacPaw says Eney already runs much of its intelligence locally, keeping reasoning, contextual search, skill execution, and conversation history on the device whenever possible. Rather than integrating another third-party model, the new partnership focuses on jointly developing the underlying AI infrastructure. After it debuts in Eney, the company plans to bring the shared models, inference framework, and memory layer to more of its own products. The same technology could eventually be made available to thousands of developers through Setapp, extending the platform beyond MacPaw's own applications. Apple has also expanded its own on-device AI tools, including the Foundation Models framework and recent demonstrations showing developers how to run AI agents locally on the Mac using MLX. The partnership reflects a broader push toward running AI workloads directly on Apple silicon while keeping more user data on the device. Apple news, rumors, tutorials, price drop alerts, in your inbox every evening, free. Unsubscribe at any time.