Private LLM Blog

Heretic vs abliterated uncensored LLMs in Private LLM: KL divergence, refusal rates, model list, and which one to run.

Run Qwen3 4B Heretic uncensored LLM on iPhone, iPad, Mac. Heretic vs abliterated: 21/100 refusals, 0.43 KL divergence, exclusive NoSlop variant.

Announcing the abliterated Qwen3 4B Instruct 2507 model in Private LLM v1.9.9 (iPhone/iPad) and v1.9.11 (Mac) - uncensored, private Local AI for NSFW roleplay, storytelling, and creative chat. Subscription Free.

Meta Llama 3.1 8B SurviveV3 is now available in Private LLM for iPhone, iPad, and Mac. This offline survival AI delivers expert wilderness guidance on modern Apple devices with 8GB RAM.

DeepSeek R1 Distill Uncensored models enable AI reasoning with NSFW and creativity on iPhone, iPad, and Mac with full privacy and offline functionality via Private LLM.

Run DeepSeek R1 Distill locally on iPhone, iPad, and Mac with Private LLM. Superior to Ollama and LM Studio, Private LLM uses advanced OmniQuant and GPTQ quantization for unmatched efficiency, accuracy, and offline privacy. Explore uncensored models and state-of-the-art reasoning with no subscriptions or cloud dependencies.

Discover the best coding AI models like Qwen 2.5 Coder for iPhone, iPad, and Mac with Private LLM. Learn how to optimize prompts, set sampling settings, and harness local AI for faster, private coding solutions

Run the best uncensored AI chat on iPhone, iPad, or Mac in 2026 with Private LLM — Heretic, Abliterated & EVA models, offline, no subscription.

Run Phi 4 locally on your Mac with Private LLM! Achieve powerful AI capabilities, including 16k context length and advanced reasoning tasks, all running locally without compromising your privacy.

Eva Qwen 2.5 is the best uncensored Qwen roleplay model on iPhone, iPad, and Mac in 1.5B, 7B, 14B, and 32B sizes. Offline, no subscription.

Private LLM now supports FuseChat 3.0 models on iPhone, iPad, and Mac. Experience faster local AI with subscription free chat using Llama 3.1 8B, Qwen 2.5 7B, and Gemma 2 9B models!

Private LLM for macOS now supports Llama 3.3 70B Euryale v2.3, unlocking advanced Llama 3.3 70B AI capabilities for creative role play and storytelling, all while ensuring your data stays private.

Private LLM for Mac now supports Llama 3.3 70B Uncensored! Experience NSFW-friendly AI chat and creative writing with this local AI solution—completely subscription free.

Private LLM for macOS now supports EVA LLaMA 3.33 70B, bringing advanced Llama 3.3 AI capabilities to your device. Dive into improved AI role play and story writing experiences, all while keeping your data secure.

Run Meta Llama 3.3 70B locally with Private LLM v1.9.4. A subscription free ChatGPT alternative for Mac users, offering fast and secure local AI processing.

Run Qwen 2.5 and Qwen 2.5 Coder models locally on iPhone, iPad, and Mac with Private LLM. Faster inference, better accuracy, and full offline functionality on Apple devices.

Run uncensored Llama 3.2 1B & 3B AI models locally on iOS and macOS with Private LLM. Experience NSFW AI responses with full privacy and subscriptions free.

Run Llama 3.2 locally on iOS: Private LLM now supports Meta's advanced multilingual AI models. Explore benchmarks, features, and how to get started.

Run Llama 3 70B locally and enjoy a powerful AI chatbot that rivals GPT-4 in performance. Private LLM offers a 4-bit quantized version of Meta's model, optimized for Apple Silicon Macs with 48GB+ RAM.

Experience the power of the uncensored Phi-3 Mini based Kappa-3 Phi Abliterated model on Private LLM. Enjoy NSFW content, roleplay, and unrestricted interactions with your AI chatbot.

Private LLM now supports DictaLM 2.0 Instruct, a bilingual (Hebrew, English) GPT chatbot. Available as a 3-bit quantized model for iPhones and iPads with 6GB+ RAM and a 4-bit model for Intel and Apple Silicon Macs. Update to v1.8.5 (macOS) or v1.8.3 (iOS) or download from the App Store for a secure, private AI experience.

Private LLM introduces Hermes 2 Pro Llama 3 8B, a new AI model based on Meta's Llama 3 8B, now available for iOS (v1.8.3) and macOS (v1.8.5). This model brings enhanced capabilities to the app, with 3-bit quantization for iOS devices with 8GB+ RAM and 4-bit quantization for Intel and Apple Silicon Macs.

Llama 3 8B Instruct model now available on Private LLM for iOS devices with 6GB+ RAM and macOS. Reduced false refusals, broader discussions, offline privacy, and coding capabilities. Update the app and download the model to get started.

Llama 3 Smaug 8B, a fine-tuned version of Meta Llama 3 8B, is now available in Private LLM for iOS and macOS. Download this model to experience an on-device local chatbot powered by Abacus.AI's DPO-Positive training approach.

Microsoft Phi-3 Mini 4K Instruct model now available on Private LLM for iPhone, iPad, and Mac. Experience superior AI performance with 4-bit quantization and 4k context length. Download the model on devices with 6GB+ RAM (iOS) or any Intel or Apple Silicon Mac and enjoy private, offline AI conversations, text summarization, and idea generation.

Private LLM introduces OpenBioLLM-8B, a cutting-edge biomedical AI model for iPhone, iPad, and Mac. Developed by Saama AI Labs, this secure, on-device chatbot delivers unparalleled performance in medical and life sciences applications. Try it today and experience the power of advanced, private AI.

Run Meta Llama 3 8B and other advanced models like Hermes 2 Pro Llama-3 8B, OpenBioLLM-8B, Llama 3 Smaug 8B, and Dolphin 2.9 Llama 3 8B locally on your iPhone, iPad, and Mac with Private LLM, an offline AI chatbot. Engage in private conversations, generate code, and ask everyday questions without the AI chatbot refusing to engage in the conversation.

Run a local LLM on iPhone, iPad, or Mac offline. 140+ models including DeepSeek R1, Llama 3.3 70B, Qwen3, and Gemma 3. No account, no subscription.

Private LLM for iOS and macOS introduces Dolphin 2.9 Llama 3 8B, an uncensored AI model that efficiently handles complex tasks like coding and NSFW conversations locally on iPhones, iPads, and Macs.

Private LLM v1.7.6 brings Google Gemma 1.1 2B and Dolphin 2.8 Mistral 7B v0.2 models to iOS devices, offering enhanced performance and uncensored AI interactions while ensuring user privacy and offline functionality. Download now and experience cutting-edge language models on your iPhone or iPad.

The latest v1.7.8 update for the macOS version of Private LLM introduces significant enhancements, notably to the Mixtral model, which now features unquantized embedding and MoE gates weights, while maintaining 4-bit OmniQuant quantization for other weights. The update also includes an increase in context length for Mistral models to 32k, provided sufficient memory is available, and introduces a grammar correction service that adapts to different English dialects. Additionally, there's experimental support for non-English European languages and a new feature allowing users to edit prompts by right-clicking. The update solidifies Private LLM's status as a top choice for running Mixtral models on Apple Silicon Macs and encourages feedback from users to improve future versions.

Explore the enhanced performance powered by Mistral-7B-OpenOrca in Private LLM v1.5 for macOS. This update brings significant memory optimization, enriching your user experience and paving the way for robust AI interactions.

Discover whats new in the v1.4 release of Private LLM for macOS. Explore OmniQuant's quantum leap in text generation, updates for WizardLM V1.2, and special options for Apple Silicon Macs. Elevate your chatbot experience with improved performance and enhancements.