Secret AI

What Secret AI does

Secret AI is a San Francisco-based, privately-held mobile software company founded in 2025 that publishes a consumer AI assistant app for iOS and Android. The app runs advanced open-source AI models — including Google Gemma 3, Meta Llama 4, DeepSeek R1, Microsoft Phi-4, Alibaba Qwen 3, OpenAI gpt-oss, and OpenBMB MiniCPM-V — entirely on-device, with no internet connection required after initial model download and no data transmitted to any server. Technical differentiation centers on supporting three inference engines (GGUF/llama.cpp, MNN from Alibaba, and Apple's MLX), direct Hugging Face model browsing, on-device vision and video understanding, full Siri Shortcuts integration, and an OpenAI-compatible custom API for connecting to self-hosted LLMs.

The business is a pure consumer, product-led-growth play. Distribution is exclusively through the iOS App Store and Google Play Store, supplemented by a content-marketing blog and a Discord community. Monetization is freemium: a Free tier with CPU inference and 15 daily messages, and Plus and Pro paid subscription tiers unlocking unlimited chat, GPU acceleration, advanced models, and vision capabilities; a one-time lifetime purchase option is also offered. Primary target users are privacy-conscious individual consumers (health, financial, legal, personal research), with secondary use cases for business professionals needing IP-protected brainstorming and travelers/remote workers needing connectivity-independent AI.

The company is bootstrapped with no disclosed external funding, operates with a team of 1–10 employees, and has executed rapid product expansion since its May 2025 launch — adding DeepSeek R1 support in June 2025 and a broader multimodal expansion (gpt-oss, LFM2 VL, enhanced vision) in August 2025. Strategic posture is privacy-first differentiation versus cloud incumbents (ChatGPT, Claude, Gemini), with the brand claim that users can verify offline operation by blocking the app's network access at the OS level.

Name

Secret AI

Legal name

Secret AI

Website

https://secretai.io

Company type

Private

Founded year

2025

Operating status

Operating

Headcount range

1–10 employees

Short description

Secret AI is a bootstrapped, San Francisco-based mobile software company founded in 2025 that publishes a freemium iOS and Android app running open-source AI models (Gemma 3, Llama 4, DeepSeek R1, gpt-oss) entirely on-device with zero data collection, serving privacy-conscious consumers, business professionals, and travelers.

Ownership category

akta.pro rank

Product category

Mobile AI Assistant Software

NAICS

Software Publishers (5132)

SIC

Services-Prepackaged Software (7372)

akta.pro primary industry

Privacy-Preserving On-Device ML (secure enclaves, TEEs, differential privacy) (HDAAAJAH)

akta.pro secondary industries

AI Privacy Engineering & Data Governance (PII, consent, retention) (HDAAAKAB), Responsible AI, Security & Privacy Platforms (Safety, Guardrails, PII) (HDAEANAG)

Keywords

• Offline AI assistant
• Mobile AI inference
• On-device AI processing
• Private AI chatbot
• Open-source AI models

Headquarters

HQ city

San Francisco

HQ country

United States

HQ region

North America

Markets served

GTM type

B2C

Offering type

Software

Cost components

Technology or R&D, Personnel, Marketing or Sales, Operations

Revenue model

1. Freemium App Model: Secret AI follows a freemium model where the app is free to download with basic features accessible at no cost. Premium capabilities including GPU acceleration for faster inference, unlimited chat, and access to advanced AI models are available through either a one-time purchase or a subscription. For in-app purchases, only official APIs from Google Play and App Store are used, with no additional servers or third-party SDKs involved.
2. Premium Model Access Upsell: Advanced models and GPU acceleration features are monetized as premium capabilities on top of the free baseline. Users can purchase either a one-time license or subscribe for ongoing access to premium model tiers and unlimited chat functionality.

Pricing tiers

Go-to-market motion1 record

Distribution channels2 records

Marketing channels3 records

Core offering

Secret AI is a mobile AI assistant app for iOS and Android that runs advanced open-source models (Gemma 3, Llama 4, DeepSeek R1, Qwen3, Phi-4, gpt-oss, MiniCPM-V) entirely offline on the user's device. All inference occurs on-device via GGUF (llama.cpp), MNN, and MLX engines with no data uploaded, no account required, and no internet connection needed after model download. Premium capabilities (GPU acceleration, unlimited chat, advanced models, vision) are offered via in-app subscription or one-time purchase on top of a free baseline tier.

Product overview

Secret AI is a mobile AI assistant application that enables users to run advanced open-source AI models completely offline on their smartphones (iOS and Android). The product centers on the Secret AI App as the core platform, which integrates multiple AI model families including DeepSeek R1, Gemma 3, Llama 4, Phi-4, Qwen3, gpt-oss, and MiniCPM-V for vision. Users download models directly from HuggingFace to their devices, and all AI processing occurs locally without internet connectivity. The system supports GGUF format models via llama.cpp, MNN format via Alibaba's mobile inference engine, and MLX for Apple devices, enabling ChatGPT-level AI interactions with complete privacy.

Differentiator

Problem solved

Functional benefit

Products and services

• Secret AI App A 100% offline, privacy-focused AI assistant mobile application for iOS and Android that runs advanced open-source AI models (Gemma 3, Llama 4, DeepSeek R1, Qwen3, Phi-4, gpt-oss, MiniCPM-V) directly on the device with no internet connection, no servers, no data collection, and no account required.

Segments3 records

Ideal customer profiles3 records

Technology focussed Yes

API detail

Has API

Yes

API docs

API detail

Core technology

AI maturity

App detail

Integration4 records

AI capability8 records

Feature6 records

Partnerships

Ten partnerships are on record, tiered core and major.

• Hugging FacecoreTechnology or IntegrationSecret AI integrates directly with Hugging Face as the primary source for AI model downloads. Users can browse and download compatible open-source models (GGUF format) directly from Hugging Face through the app's interface, with no middleman or markup. This is a foundational technical integration essential to the app's value proposition.
• Google (Gemma models)coreTechnology or IntegrationSecret AI supports Google's Gemma 3 open-source models, enabling users to run Google-quality AI entirely offline on their mobile devices. Gemma 3 provides state-of-the-art vision-language capabilities including text and image understanding.
• Meta (Llama models)coreTechnology or IntegrationSecret AI integrates Meta's Llama 4 open-source models, providing access to Meta's flagship open AI models for offline mobile use. Models include Llama 4 variants with Dolphin fine-tunes available.
• DeepSeek (DeepSeek R1 models)coreTechnology or IntegrationSecret AI supports DeepSeek R1 reasoning models, including the DeepSeek-R1-0528 variant with Qwen3 and Llama base models. DeepSeek R1 is highlighted as a flagship model for offline reasoning tasks with chain-of-thought capabilities.
• Alibaba (Qwen models and MNN engine)coreTechnology or IntegrationSecret AI integrates Alibaba's Qwen3 models for multilingual AI and uses Alibaba's MNN inference engine for optimized mobile inference performance. Qwen3-2507 provides advanced reasoning with 100+ language support. MNN is a lightweight high-speed inference engine specifically designed for mobile and embedded devices.
• Apple (MLX engine, iOS/macOS integration)majorTechnology or IntegrationSecret AI leverages Apple's MLX inference engine for optimized performance on Apple devices. Full Siri Shortcuts integration is provided for iOS and macOS. The app supports macOS 13.3 or later and is optimized for Apple hardware with MLX format models.
• OpenAI (gpt-oss models, API compatibility)majorTechnology or IntegrationSecret AI supports OpenAI's gpt-oss open-weight models and maintains OpenAI-compatible API support, allowing users to connect to their own self-hosted LLM API services while maintaining local privacy.
• Microsoft (Phi models)majorTechnology or IntegrationSecret AI supports Microsoft's Phi-4 model, an efficient state-of-the-art open model built on synthetic and high-quality public data, optimized for reasoning and logic tasks in memory-constrained environments.
• llama.cpp (GGUF inference engine)coreTechnology or IntegrationSecret AI uses llama.cpp's GGUF format as one of its three primary AI inference engines. GGUF provides the largest community ecosystem and most rapid model support, enabling the broadest model compatibility across the app.
• OpenBMB (MiniCPM models)majorTechnology or IntegrationSecret AI supports OpenBMB's MiniCPM-V 4.5 and MiniCPM-o 2.6 vision-language models, providing GPT-4o-level multimodal capabilities for image understanding and video analysis entirely on-device.

Scale indicators5 records

Recent moves6 records

Expansion highlights6 records

Emerging players

• ChatterUI: Mobile AI chat app focused on running open-source models directly on the device. Overlaps with Secret AI on the core offline-on-mobile use case, though typically positioned for more technical users.

Broad incumbents

• Apple Intelligence: Apple's system-level on-device AI features integrated across iOS, iPadOS, and macOS. A broad incumbent that threatens Secret AI's standalone position by delivering private AI as part of the OS itself.
• Google Gemini Nano: Google's smallest Gemini variant, designed for on-device execution on Pixel and select Android devices. Competes directly with Secret AI's on-device inference value proposition, embedded inside the Android platform.

Direct peers

• Ollama: Open-source local LLM runner enabling users to run open-weight models offline on their own hardware. Closest comparable on the 'private, self-hosted AI assistant' value proposition; differs in being CLI/desktop-first rather than mobile consumer.
• PocketPal AI: Free, open-source mobile app that runs small language models directly on iPhone and Android. Nearly identical consumer proposition — offline open-model chat on phone — and the most direct mobile-app competitor.
• LM Studio: Desktop GUI for running local LLMs privately. Shares the on-device/open-model philosophy and target user (privacy-conscious power users) but is focused on macOS/Windows rather than mobile.
• MLC AI: Open-source project (MLC LLM) that compiles and optimizes LLMs for on-device deployment including mobile. Tackles the same technical problem of running open models locally and is a direct infrastructure-level peer.
• PrivateGPT (by Zylon): Privacy-focused local AI for document question-answering. Same 'AI without sending data to the cloud' positioning as Secret AI but oriented toward enterprise document search rather than a general mobile assistant.
• Maid (Mobile Artificial Intelligence Distribution): Mobile-focused local AI app running models on-device without servers. Shares the 'private offline AI on phone' positioning aimed at privacy-conscious consumers.

Others

• Hugging Face: Central infrastructure provider for open-source AI model distribution. Secret AI is built directly on Hugging Face's model hub; not a direct competitor but the critical enabling dependency for the product.

Market position

Strengths4 records

Weaknesses4 records

Competitive moat4 records

Key risks5 records

Key highlights5 records

Customer concentration

• Xx.com/SecretAILabs
• LinkedInlinkedin.com/company/secret-ai-labs

Revenue estimate

Valuation estimate

Number of profiles

Funding overview

Funding rounds

Investors

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

M&A

Investments

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

Frequently asked questions about Secret AI