Is Qwen3.8-Max available as open weights?

Yes. Qwen3.8-Max was released on August 3, 2026 as the first Max-class Qwen model to ship with open weights. You can run it via providers like OpenRouter, Together AI, or the official Qwen API. A locally-runnable Qwen3.8-27B open-weight variant is also planned for the following week, which will be practical to run on a high-end consumer GPU.

How does Qwen3.8-Max compare to Claude or GPT for coding?

Qwen3.8-Max is positioned as a strong coding and collaboration model, with Alibaba claiming it sets a new bar in that category. In practice, Claude Opus 5 (released July 24, 2026) still leads on complex multi-file code generation and architecture reasoning. Qwen3.8-Max is competitive for single-file tasks, has strong multilingual coding support, and its open-weight status means it can be self-hosted or fine-tuned, which closed models cannot match.

How do I build an app using Qwen3.8-Max without managing infrastructure?

Use the Qwen API or OpenRouter (which added Qwen3.8-Max on day one) as the model endpoint, and build the product layer with an AI builder. With We.Inc you describe the app in plain English, get a live working React app immediately, wire in your Qwen API key as the model call, and publish to a custom domain. The generated code is yours to export, so you get the same no-lock-in ownership that makes open weights attractive. Start free at we.inc.

Can I run Qwen3.8-Max locally?

Qwen3.8-Max itself is a large model not designed for consumer local inference. However, Alibaba has announced Qwen3.8-27B as a separate open-weight release planned for the week of August 10, 2026. The 27B size runs on a high-end GPU (an RTX 4090 with 24GB VRAM is sufficient for quantized inference) and carries much of the Max's capability for coding tasks. Until then, the Qwen API and OpenRouter are your fastest on-ramps.

Qwen3.8-Max Just Dropped as Open Weights. Here Is What Developers Are Building Today.

Alibaba released Qwen3.8-Max on August 3, 2026 -- the first time a Max-class (frontier-tier) Qwen model has shipped with open weights. The 27B local-runnable version follows next week. This guide covers what changed, how Qwen3.8 benchmarks compare, and the fastest path to shipping a real product on top of it today.

Alibaba launched Qwen3.8-Max on August 3, 2026, and within hours it was front-page on Hacker News with nearly 1000 points. The headline is not just another model drop: this is the first time Alibaba has shipped a Max-class model -- their frontier tier -- with open weights. OpenRouter had it live on day one. A locally-runnable Qwen3.8-27B follows next week.

**TL;DR:** Qwen3.8-Max is now the strongest open-weights coding model available via API. If you want to build a product on top of it today, the Qwen API or OpenRouter gets you running in minutes, and you can wrap a full working app around that endpoint using [We.Inc](/) without writing the infrastructure layer. The 27B local-runnable variant drops next week and will run on a 24GB GPU.

What Changed in Qwen3.8

The Qwen line has always been strong for its size class. Qwen 3.6-27B earned a reputation as one of the best local models across coding, reasoning, and multilingual tasks, and nothing close matched it at that size. Qwen3.8 is the jump to the Max tier, and the relevant facts are:

**Open weights at frontier grade.** Previous Qwen-Max releases were closed API only. Releasing weights at this level of capability is the headline, not a benchmark number. It means you can self-host, fine-tune, or run it behind your own firewall without depending on Alibaba's API staying up.

**Coding and "cowork" is the stated lead use case.** The official blog positions Qwen3.8-Max as a model built for software development workflows where a developer and an AI exchange context over a long session: not a one-shot autocomplete but an ongoing pair-programmer. Benchmarks released with the model show it at or above previous Qwen-Max versions on HumanEval, MBPP, and multi-turn coding scenarios.

**27B local variant confirmed for next week.** The HN thread confirmed a Qwen3.8-27B release coming the week of August 10. Qwen3.6-27B is already widely regarded as the best local model in that size class, and the 27B variant of 3.8 will be the one developers run on their own machines.

**OpenRouter on day one.** This matters practically: if you already have an OpenRouter key, you can call Qwen3.8-Max right now with zero provider setup.

How Qwen3.8-Max Fits the Current Landscape

Three releases shaped the frontier in the last two weeks: Claude Opus 5 (July 24), Kimi K3 open weights (late July, 2.8T parameters), and now Qwen3.8-Max (August 3). They serve different niches:

| Model | Best For | Open Weights | Runnable Locally | |---|---|---|---| | Claude Opus 5 | Complex multi-file generation, architecture | No | No | | Kimi K3 | Long-context tasks, research docs | Yes | No (2.8T params) | | Qwen3.8-Max | Coding, multilingual, long sessions | Yes | Via API now, 27B local next week | | Qwen3.8-27B (coming) | Local coding assistant | Yes | Yes, 24GB GPU |

If you want to ship a product that calls a frontier model and you care about open weights, Qwen3.8-Max is now the most practical choice. Kimi K3 is open weights too, but at 2.8 trillion parameters it requires datacenter hardware to run. The Qwen3.8-27B will be the first frontier-grade model that a developer can actually run on a workstation.

What People Are Building With It Today

The practical build cases showing up in the early Qwen3.8 thread:

**Long-session coding assistants.** The "cowork" positioning is real: Qwen models handle long context well, which makes them suited for sessions where you are building a feature across dozens of files rather than asking one-shot questions. Developers at smaller companies who want an internal AI pair-programmer they can self-host are the target.

**Multilingual SaaS tools.** Qwen has historically been the strongest model family for non-English coding contexts, particularly CJK languages. Teams building tools for Japanese, Chinese, or Korean markets are already adopting it as a drop-in replacement for English-first models.

**Fine-tuned internal tools.** Open weights means fine-tuning is on the table. Companies building specialized coding agents for their own stacks (proprietary SDKs, internal APIs, domain-specific frameworks) can tune Qwen3.8 on their codebase in a way that closed models do not allow.

The Fastest Path to Shipping a Qwen3.8-Max App

The model endpoint is the easy part. The product around it is where projects stall. Here is the concrete path using [We.Inc](/):

1. **Get a Qwen API key** at qwen.ai, or use OpenRouter if you already have an account there. You get a URL and a key -- the same interface as any OpenAI-compatible endpoint.

2. **Describe the product in plain English.** Open We.Inc and type what you are building: "A pair-programmer UI where a developer can paste a file, ask questions about it, and get a response from a coding AI. Clean dark interface with a code editor on the left and chat on the right." We.Inc generates a live, working React app you can click in seconds.

3. **Wire in your Qwen endpoint.** In the generated code, point the model call at your Qwen API URL with your key. If the app needs a different prompt or system message, edit it inline or tell the AI to change it.

4. **Publish and optionally export.** Publish to a custom domain from the builder, or export the full React codebase to your GitHub. Because you own the generated code, the same open-weights principle that brought you to Qwen applies to your build tool: no lock-in.

The typical time from "describe the app" to "live on a domain" is under an hour for a single-feature tool. The infrastructure, auth wiring, and hosting that would normally be the week-long part are handled.

Qwen3.8 vs Claude for Building With WeInc

One question worth addressing directly: WeInc uses Claude models for the AI generation inside the builder itself. Does that conflict with building a Qwen-powered app on top of WeInc?

No, and this is the normal architecture now. The AI that writes and edits your code (the builder AI) is separate from the AI your app calls at runtime (the product AI). We.Inc uses Claude because it currently leads on multi-file code generation and UI reasoning. The app WeInc generates can call any API you specify -- Qwen3.8-Max, a local Qwen3.8-27B instance, Claude, or GPT-4o. You are picking the best tool at each layer, not committing to one model for everything.

Start Building

Qwen3.8-Max is live on OpenRouter right now. The 27B local variant follows next week. The window where being early on a model gets you distribution -- citations, links, word of mouth -- is roughly the first two weeks after launch.

Describe your Qwen-powered app at [We.Inc](/), free to start with no card required, and have it live on a domain today. If you want to compare plans first, pricing is at [/pricing](/pricing). For a broader look at building production apps without code, see [how to build a web app without code](/blog/how-to-build-a-web-app-without-code-2026).

More in Blog

We.Inc is an AI-powered website builder you can resell under your own brand. Launch a branded client dashboard, bill on Stripe Connect, and deliver AI-generated websites in minutes. White-label plans from $499/mo, no per-site fees.

Product

AI Website Builder: Describe a site in plain English; it's live in minutes
White Label Platform: Your brand, your domain, your margin
Enterprise: SSO, custom SLAs, and dedicated support
Developer API: Embed site generation into your SaaS

Who It's For

For Agencies: Resell under your brand
For Freelancers: Your own branded builder
For Hosting Providers: Add a builder to your stack
For SaaS Platforms: Embed site creation via API
For Startups
For Financial Services
For Healthcare

Features

Resources

Company

View Sitemap