OfflineLLM: Private AI Chat

About this app

<p>OfflineLLM uses a custom execution engine for LLMs, optimised specifically for Apple Silicon on iPhone, iPad, Mac and Vision Pro. Using the full power of Metal 3, OfflineLLM is able to run faster on consumer devices than apps based on llama.cpp and MLC.</p>
<p>Welcome to AI chatbots that can be used without the risk of compromising confidential and sensitive data.</p>
<p>For the first time, you can have personal GPT assistants running privately on your device without an Internet connection. No data is ever sent to the cloud. Your conversations never leave your device. Continue using your AI chatbots when in Airplane mode.</p>
<p>Chat directly with the LLM that powers Apple Intelligence!</p>
<p>Experience multi-modal vision models like never before on the iPhone. Send images to your offline AI chatbots.</p>
<p>Further enhance your AI chatbots using RAG (Retrieval Augmented Generation) to integrate your own documents and files into your LLMs.</p>
<p>Beginner mode allows novices to step into the world of LLMs and AI chatbots.</p>
<p>Advanced mode allows experts to configure every parameter of the LLM execution engine. </p>
<p>Interact with Offline LLM, anywhere and on any device. Extend the app with system Shortcuts, Automations and Widgets.</p>
<p>Have other tools? Use our OpenAI-compatible API to integrate with other systems.</p>
<p>Use state-of-the-art AI models, such as DeepSeek, Llama or Gemma to privately answer any questions you may have. Offline LLM supports any GGUF model, including:<br />
- GPT-OSS<br />
- DeepSeek<br />
- Llama<br />
- Gemma<br />
- Phi<br />
- Mistral<br />
- ORCA<br />
- MobileVLM<br />
- StableLM<br />
- Code Llama<br />
- VisionLlama<br />
- TinyLlama<br />
- OpenHermes<br />
- WizardLM<br />
- RWKV</p>
<p>Whether you are getting ready for an important presentation, a copywriter looking for a clever turn of phrase, or a student needing to summarise a large article of text for an essay, an AI chatbot can help you. Offline LLM can be your personal AI assistant.</p>
<p>AI Writing Assistant:<br />
Get personalised help from your AI chatbot. Draft anything from emails and speeches to lyrics and poems. </p>
<p>Grammar and Spelling Checker:<br />
Using the fastest GPT on the App Store, quickly check and correct your grammar, spelling and function.</p>
<p>Professional Proofreading and Rewriting:<br />
Use AI to proofread and rewrite your text to make it more engaging, coherent and professional.</p>
<p>Summarise Text:<br />
Let your personalised AI assistant read large articles of text and summarise them neatly and concisely so that you can quickly understand large chunks of information in record time.</p>
<p>Personal Tutor:<br />
Learn any concept quickly and easily. Ask your GPT tutor to explain any concept for you to learn in record time. Tackle homework and assignments with ease and get top marks.</p>
<p>Coding Assistant:<br />
Your personalised AI coding assistant. Get help programming in any language, including Python, JavaScript, TypeScript, Java, Swift, Kotlin, PHP, Go, C/C++, Haskell, Perl, Ruby, Rust, C# and more. Write efficient and reliable production code in record time with the help of AI.</p>
<p>Learn Languages:<br />
Learn any language with the help of a fluent AI chatbot. Master any language in record time.</p>
<p>Ultimate Creativity Expert:<br />
Get instant creativity at your fingertips with the help of an AI creative. Impress others with your newfound creativity.</p>
<p>AI Friend:<br />
Chat with a companion 24/7 with access to an AI chatbot. Quick answers in real-time to all of your messages. No more being left on blue ticks.</p>
<p>Gourmet Chef:<br />
Get delicious food recipes and meal inspiration from an AI expert.</p>
<p>Offline LLM can run a variety of models, which are capable of running on a range of devices. Performance on devices will vary and depend on the model selected. More advanced models are optimized for powerful Macs and iPads and will struggle to run on older iPhones. To successfully run a model, your device must have more VRAM than the model you choose to use. Large language models perform differently for each question asked; some questions will run significantly quicker or slower than others.</p>