nostr relay proxy

event page

# Best Local LLMs For Android 👇 I have been testing the latest models on my **Google Pixel 7 Pro with GrapheneOS** and the [PocketPal](https://github.com/a-ghorbani/pocketpal-ai) app. Here are my recommendations. --- ## Qwen 3 4B - đŸ”—ī¸ [`unsloth/Qwen3-4B-GGUF`](https://huggingface.co/unsloth/Qwen3-4B-GGUF) - đŸ­ī¸ Made by Alibaba Cloud - đŸ“†ī¸ Released in April 2025 - đŸ’Ąī¸ First all-rounder model with a switch for thinking and non-thinking mode - đŸ§Ēī¸ Use it for: - Math - Logic - Coding - Translations - Content creation - Summarizations - Conversions to structured data - â„šī¸ To turn on thinking mode, append `/think` to your prompt - `How many 'r's are in 'strawberries'? /think` - â„šī¸ To turn off thinking mode, append `/no_think` to your prompt - `Who are you? /no_think` - â„šī¸ For math questions, append `Please reason step by step, and put your final answer within \boxed{}.` to your prompt - `At a restaurant, each adult meal costs $5 and kids eat free. If a group of 15 people came in and 8 were kids, how much would it cost for the group to eat? Please reason step by step, and put your final answer within \boxed{}.` - â„šī¸ Recommended settings for thinking mode - Temperature: `0.6` - Top P: `0.95` - Top K: `20` - Min P: `0` - â„šī¸ Recommended settings for non-thinking mode - Temperature: `0.7` - Top P: `0.8` - Top K: `20` - MinP: `0` --- ## Gemma 3 4B - đŸ”—ī¸ [`MaziyarPanahi/gemma-3-4b-it-GGUF`](https://huggingface.co/MaziyarPanahi/gemma-3-4b-it-GGUF) - đŸ­ī¸ Made by Google - đŸ“†ī¸ Released in March 2025 - đŸ’Ąī¸ Powerful vision language model with support for many languages - đŸ§Ēī¸ Use it for: - Translations - Content creation - Summarizations - Conversions to structured data - Code creation - Language learning ## DeepSeek R1 Distill Llama 8B - đŸ”—ī¸ [`lmstudio-community/DeepSeek-R1-Distill-Llama-8B-GGUF`](https://huggingface.co/lmstudio-community/DeepSeek-R1-Distill-Llama-8B-GGUF) - đŸ­ī¸ Made by DeepSeek - đŸ“†ī¸ Released in January 2025 - đŸ’Ąī¸ Great reasoning model - đŸ§Ēī¸ Use it for: - Math - Logic - Coding - â„šī¸ Avoid adding a system prompt and use user prompt only - â„šī¸ Recommended temperature setting is `0.6` - â„šī¸ For math questions, append `Please reason step by step, and put your final answer within \boxed{}.` to your prompt - `At a restaurant, each adult meal costs $5 and kids eat free. If a group of 15 people came in and 8 were kids, how much would it cost for the group to eat? Please reason step by step, and put your final answer within \boxed{}.` --- ## EXAONE Deep 7.8B - đŸ”—ī¸ [`LGAI-EXAONE/EXAONE-Deep-7.8B-GGUF`](https://huggingface.co/LGAI-EXAONE/EXAONE-Deep-7.8B-GGUF) - đŸ­ī¸ Made by LG - đŸ“†ī¸ Released in March 2025 - đŸ’Ąī¸ Underrated reasoning model from Korea - đŸ§Ēī¸ Use it for: - Math - Logic - Coding - â„šī¸ Use only English or Korean prompts - â„šī¸ Avoid adding a system prompt and use user prompt only - â„šī¸ Recommended settings - Temperature: `0.6` - Top P: `0.95` #android #ai #llm #privacy

rendered in 6.878117ms