# Strata review

> Strata is a local inference engine and app that runs large open models like Qwen3.8-Flash-Next on consumer hardware with OpenAI/Anthropic APIs.

Devtoolwatch Board, the week of Oct 25–31, 2026. Canonical page: https://index.devtoolwatch.com/tools/niko1221-strata/

## TL;DR

Strata runs models such as Qwen3.8-Flash-Next, Qwen3.6-35B-A3B and DeepSeekV4.1-Flash on your own computer. Strata is for developers who want local chat and coding without an account. It offers a desktop app, a terminal app, and OpenAI- and Anthropic-compatible APIs on localhost for tools like Claude Code and Codex CLI.

## Strata at a glance

- **Category:** [Inference and training](https://index.devtoolwatch.com/categories/inference-and-training/)
- **Maintained by:** Niko Veit (individual, [@Niko1221](https://github.com/Niko1221) on GitHub)
- **License:** MIT
- **Language:** C++, Python, Cuda
- **Repository:** [github.com/Niko1221/Strata](https://github.com/Niko1221/Strata)
- **Latest release:** [v0.2.9](https://github.com/Niko1221/Strata/releases/tag/v0.2.9) on GitHub, Oct 31, 2026
- **First published:** Sep 24, 2026

## What Strata does

- Strata runs Qwen3.8-Flash-Next, Qwen3.6-35B-A3B and DeepSeekV4.1-Flash locally, with a choice of quantized model packs from several providers.
- Strata serves OpenAI- and Anthropic-compatible APIs on localhost, with optional image input.
- Strata connects to coding tools such as Claude Code, Codex CLI, OpenCode and Hermes Agent through its compatible APIs.
- Strata's desktop app combines chat, files and a live preview in one window.
- Strata reads DeepSeek model weights that do not fit in memory from your SSD.
- Strata's terminal app lets you watch your model and hardware while it runs.

### How it works

Strata runs as a desktop app or a terminal app. On first start the desktop app downloads the engine for your graphics card. In the terminal, strata launch prompts you to download and start a model. The engine exposes OpenAI- and Anthropic-compatible APIs on localhost, and strata connect shows setup instructions for coding tools. Docker and building from source are also documented.

## How to install Strata

**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
```

### Requirements

- A supported GPU or system: NVIDIA RTX 20–50 series, supported AMD Radeon cards, Intel Arc on Linux, Apple Silicon M1 or newer, Strix Halo, or DGX Spark.
- DeepSeekV4.1-Flash needs about 245 to 280 GB of free disk space, depending on the set.
- DeepSeekV4.1-Flash on Apple Silicon needs a Mac with 64 GB or more and is text only.
- Strata needs no account.

### First steps

1. Install the terminal app. On Windows, open a new terminal afterwards.
2. Run strata launch and follow the prompts to download and start a model.
3. Run strata connect to see setup instructions for your coding tool.

## Who Strata is for

- Developers who want to run large open models on their own consumer GPU.
- Users of Claude Code, Codex CLI or OpenCode who want a local model backend.
- Developers who want a local, OpenAI- or Anthropic-compatible API endpoint.

### Limitations

- Strata's AMD Radeon RX 6800 / 6900 support is experimental, and Intel Arc and DGX Spark support is Linux only.
- Strata's DeepSeek sets are 239 GB or larger to download and rely on SSD reads when memory runs short.
- Strata's Gyro model packs use less memory but run slower.

## Usage data (the week of Oct 25–31, 2026)

- Board rank: #4 overall, #3 in Inference and training, momentum score 98
- New GitHub stars per week (last 12 weeks): Aug 9: n/a; Aug 16: n/a; Aug 23: n/a; Aug 30: n/a; Sep 6: n/a; Sep 13: n/a; Sep 20: 217; Sep 27: 8.5k; Oct 4: 12k; Oct 11: 6.1k; Oct 18: 5.2k; Oct 25: 5.1k

## Strata alternatives

| Rank | Tool | Stars | Downloads/wk |
|---|---|---|---|
| #1 | [dsh-our-free-model](https://index.devtoolwatch.com/tools/ebony-vinyl-dsh-our-free-model/) | 54.3k | – |
| #2 | [magpie](https://index.devtoolwatch.com/tools/yetone-magpie/) | 26.9k | – |
| #4 | [edge0](https://index.devtoolwatch.com/tools/edge0-ai-edge0/) | 8.9k | – |
| #5 | [OmniRoute](https://index.devtoolwatch.com/tools/diegosouzapw-omniroute/) | 83.2k | 106k |
| #6 | [CLIProxyAPI](https://index.devtoolwatch.com/tools/router-for-me-cliproxyapi/) | 57.1k | – |
| #7 | [codex-chatgpt-web](https://index.devtoolwatch.com/tools/miuuyy-codex-chatgpt-web/) | 15.3k | – |

## Strata FAQ

### Is Strata free and open source?

Yes. Strata's source code is on GitHub (Niko1221/Strata) under the MIT license, so you can use it for free.

### How do I install Strata?

Run `curl -fsSL https://github.com/Niko1221/Strata/releases/latest/download/install.sh -o install.sh && sh install.sh`. The install section above has the other options from the project's README.

### Is Strata still maintained?

Yes. The Niko1221/Strata repository was last updated on Oct 31, 2026, and the latest release, v0.2.9 on GitHub, came out on Oct 31, 2026.

### How popular is Strata?

Strata has 37.1k GitHub stars as of Nov 1, 2026. It gained 5.1k new stars in the week of Oct 25–31, 2026, against a 12-week average of 6.2k a week. That made it #3 among LLM inference and training tools on the Devtoolwatch Board.

### What are the best alternatives to Strata?

Other LLM inference and training tools on the Devtoolwatch Board in the week of Oct 25–31, 2026, by momentum: dsh-our-free-model (#1), magpie (#2), edge0 (#4), OmniRoute (#5), CLIProxyAPI (#6). The alternatives table above has their numbers.

### Which coding tools work with Strata?

Strata works with Claude Code, Codex CLI, OpenCode, oh-my-pi, pi.dev, Hermes Agent, OpenClaw, and other tools through its OpenAI- and Anthropic-compatible APIs. Run strata connect to see setup instructions for your tool.

### Does Strata need an account or an internet API?

Strata needs no account. Models run on your own computer, and Strata serves its APIs on localhost.

### Which model pack should I pick in Strata?

Strata recommends ISTA-DASLab packs first, then Unsloth and other providers where a compatible model fits. For coding, the README points to a Coder IQ1_M pack. A guide in docs/MODELS.md helps you choose by memory.

### Can Strata run DeepSeekV4.1-Flash on a Mac?

Strata runs DeepSeekV4.1-Flash on Apple Silicon Macs with 64 GB or more, text only. Support was added in release 0.2.7.1.

### Does Strata accept images?

Strata supports optional image input.

## Sources

1. [Niko1221/Strata on GitHub](https://github.com/Niko1221/Strata)
2. [Strata README](https://github.com/Niko1221/Strata/blob/main/README.md)
3. [Release v0.2.9 (Oct 31, 2026)](https://github.com/Niko1221/Strata/releases/tag/v0.2.9)
4. [Release v0.2.8 (Oct 29, 2026)](https://github.com/Niko1221/Strata/releases/tag/v0.2.8)
5. [Release v0.2.7 (Oct 27, 2026)](https://github.com/Niko1221/Strata/releases/tag/v0.2.7)
6. [Hacker News: Run Qwen 3.8 Flash Next (125B) on consumer hardware (RTX 4090) at 100T/s](https://news.ycombinator.com/item?id=49953495)
7. [Hacker News: Strata: Qwen3.8-Flash-Next (125B Moe) on a 8GB+ Nvidia GPU](https://news.ycombinator.com/item?id=49879253)
8. [Devtoolwatch Board, the week of Oct 25–31, 2026](https://index.devtoolwatch.com/)