Run powerful AI on your own computer.

Chat, code and work with images, locally.

Models · Install · Hardware · Tools · Help · Support

_{English · 简体中文 · 日本語 · Deutsch · Français · Español · Português}

Free, open source, and yours to run. Use Strata on your desktop or with your favorite coding tools. No account needed.

Personal fork (c-OOL-io/Strata-WebSearch). This fork adds web search and full page reading to the built-in Chat and the serve frontend: the lean terminal web app, where upstream has no search (upstream's search lives in the App). It is built the way Strata's Chat itself is built: completely off by default (until you turn it on, nothing search-related leaves the PC), no added tokens for chats that don't enable it, and search text budget-capped to a small slice of the model's context window (defaults scale with max_context, so small-context instances stay safe).

• Engines: DuckDuckGo by default (no key), SearXNG (your own self-hosted instance), Brave, Tavily, Serper (Google), Exa and Perplexity with your own keys. Keys stay in the browser; nothing search-related is stored on the server.
• Query optimizer: conversational wording is rewritten into a clean search query by one fast call to your own local model, kept in the language you wrote in. If it fails, the raw message is searched instead. Whenever what was searched differs from what you typed, the Sources line shows it.
• Read result pages: the top result pages are fetched and injected as extracted text (nav/footer/scripts stripped, main/article preferred), with Direct / Jina-fallback / Jina-for-all reader modes and an explicit warning whenever a third party could see your URLs.
• Auditable answers: every reply gets a Sources line (engine, exact query, clickable links, read badges) that rides along into the Markdown export, and a failed search leaves a named per-cause warning (missing key, unreachable engine, bot check) instead of failing silently.
• Hardened: /search sits behind the app's own-page guard (foreign origins rejected), http(s)-only URL checks so page reading can never become a local file reader, injected web text is framed as untrusted, and the backend (serve/websearch.py) is stdlib-only, no new dependencies, Windows/Linux/macOS.

Two things to know before you download anything:

1. The download links below are the vanilla upstream app from Niko1221/Strata. The prebuilt binaries do not contain this fork's search. To run the search-enabled version, use this repo's source: git clone, branch web-search, then START-HERE.bat / setup.sh.
2. This fork is also for the resource-light Terminal route: anyone who only wants the lean serve/CLI on a headless server (no app, no chat UI) gets it from this same source, with the search built into the serve frontend.

Install via CLI (this fork, branch web-search): clone and set up exactly as you would Strata from source, with the search built in. The fork's own install guide (terminal / from source, and how to update) is in docs/INSTALL.FORK.md:

git clone -b web-search https://github.com/c-OOL-io/Strata-WebSearch.git && cd Strata-WebSearch
./setup.sh          # first run: .venv, the engine build, the model; then it serves

Bench results from an RTX 4080 SUPER / Ryzen 9 7950X3D are included in bench/.

_{Left: an answered chat with the Sources line under the reply (engine used, the exact query that was searched). Web search switches on/off via the globe in the composer.
Right: the Settings panel's Web search section: on/off, query optimization, engine picker (DuckDuckGo / SearXNG / Brave / Tavily / Serper / Exa / Perplexity), key field, page reading.}

Models

We recommend ISTA-DASLab first, then Unsloth and other providers, where a compatible model fits.

 Qwen3.8-Flash-Next

Provider	Choices
ISTA-DASLab · recommended	Q2_0, IQ2_XS, IQ3_XXS, IQ3_S. Coder: IQ1_M for coding.
Unsloth	UD-IQ4_XS, UD-Q4_K_XL, UD-Q5_K_XL, UD-Q6_K_XL, Q8_0
Gyro · agentionai	Gyro-S, Gyro-M. Lower memory use but slower.
Other options	Swift 1.5, NVFP4, OrcaRouter

Compare sizes, memory needs and compatibility.

 Qwen3.6-35B-A3B

Unsloth: UD-Q4_K_XL, UD-Q6_K_XL, Q8_0. These are the currently documented packs for this model.

Model guide and compatibility.

 DeepSeekV4.1-Flash-1017

In the desktop app and the Terminal. What does not fit in memory is read from your SSD, so it needs a lot of free disk space.

Set (IST-DAS)	Download	Free disk needed
1.75 bpw	239 GB	about 245 GB
2.00 bpw	257 GB	about 265 GB
2.25 bpw	274 GB	about 280 GB

Runs on NVIDIA RTX 30, 40 and 50 series (Windows and Linux) and AMD Radeon AI PRO R9700 (Linux). Setup downloads the files and the DeepSeek engine, then starts the model.

strata launch --family deepseek41 --model ist-das-2.00bpw --plan

Measured with the 2.00 bpw set on Strata 0.2.5. Output is the speed of the answer, prompt is how fast Strata reads your input.

Hardware	Output	Prompt
RTX 5070 12 GB, 64 GB RAM	4 to 6 tokens/s	about 110 tokens/s (4K)
RTX 3060 12 GB, 46 GB RAM	9 to 11 tokens/s	about 155 tokens/s (4K)
1x Radeon AI PRO R9700, 251 GB RAM	21 to 25 tokens/s	320 to 360 tokens/s (8K)
4x Radeon AI PRO R9700	74 to 77 tokens/s	400 to 450 tokens/s (8K)

Memory, disk and details · Release notes

Not sure what fits? Choose a model for your computer.

Install

Desktop app

Download Strata for Windows, Linux or Mac →

Install the app, choose a model, and start chatting. There is one installer for each system. On first start the app downloads the engine for your graphics card. Help with the installer.

_{Files: Strata-0.2.6-windows-x64-setup.exe · Strata-0.2.6-linux-x86_64.AppImage · Strata-0.2.6-macos-arm64.dmg}

_{Chat, files and a live preview in one window.}

Terminal app

Prefer the terminal? Install Strata, then run strata launch.

Install commands

Linux / Mac

curl -fsSL https://github.com/Niko1221/Strata/releases/latest/download/install.sh -o install.sh && sh install.sh

Windows PowerShell

irm https://github.com/Niko1221/Strata/releases/latest/download/install.ps1 | iex

On Windows, open a new terminal afterwards. Then:

strata launch

Follow the prompts to download and start a model. Terminal guide.

_{Watch your model and hardware while it runs.}

Build from source · Docker · Full install guide

Hardware

Supported hardware	Guide
NVIDIA GeForce RTX 20 / 30 / 40 / 50 series	NVIDIA setup
AMD Radeon RX 7700 / 7800 / 7900, RX 9060 / 9070, AI PRO R9700	AMD setup
AMD Radeon RX 6800 / 6900 · experimental	RX 6000 setup
Intel Arc · Linux	Intel setup
Apple Silicon · M1 or newer	Mac setup
AMD Ryzen AI Max · Strix Halo	Strix Halo setup

Tools

Claude Code · Codex CLI · OpenCode · oh-my-pi · pi.dev · Hermes Agent · OpenClaw, and more through the OpenAI- and Anthropic-compatible APIs.

Run strata connect to see setup instructions for your tool. Connection guide.

Help

All guides · Troubleshooting · Report a bug · Benchmarks · How it works · Contribute

Support Strata

Help keep local AI open to everyone. A coffee helps fund development and testing on more hardware.

MIT license · Credits and model licenses · Privacy

_{For AI tools and automated readers: technical reference · setup instructions.}

Sponsorships, hardware partnerships or collaborations: niko.veit@outlook.com.