Qwen 3.8 Max quietly appears on qwen.ai

Models & Tools il y a 10 h7Ajouter aux favoris

Qwen 3.8 Max quietly appears on qwen.ai
Illustration : Léa Fontaine

Alibaba's flagship ticks past 3.5. No blog post yet - just a product page and a version bump.

In plain terms

Alibaba's top-tier Qwen model has ticked from 3.5 to 3.8 Max on qwen.ai. No benchmarks, no paper, no launch post - just the model appearing on the homepage.

The story

The Qwen 3 line has been Alibaba's most credible answer to Kimi K3 and Claude Fable 5 through the first half of 2026. Version 3.5-122B has been drilled into by the developer community (we covered its Mac Studio daily-driver take on 12 July). What's new is that a "3.8 Max" tier is now live on the qwen.ai homepage without the usual launch playbook. It fits a broader pattern at Alibaba of shipping model updates ahead of formal documentation - but the specifics of previous version jumps, and how quickly the paperwork followed, are not something we'll relitigate here.

Under the hood

No parameter count, no context window, no MMLU/HumanEval numbers publicly released as of writing. The Max suffix historically indicates Alibaba's closed-weights tier - distinct from the open-weight Qwen3-*B releases.

So what

Watch for a technical report and third-party benchmarks in the next 72 hours. Until then, treat "3.8 Max" as a claim: Alibaba is shipping fast enough that its version numbers now outpace its documentation. For teams already running Qwen in production, the update is worth queueing for evaluation - not for immediate switch-over.

Ressources, à tester

Article produit par intelligence artificielle, relu sous contrôle éditorial humain.

Notre rédaction
Your Linux servers, as a desktop.
TermalOSSponsored
Ops, reimagined

Your Linux servers, as a desktop.

Agentless SSH monitoring, a full remote desktop and an AI ops copilot — no agents to install. Everything stays on your machine.

SSHMonitoringAI Ops
Get early access
Cet article vous a-t-il été utile ?

10 personnes ont aimé cet article

J'aime
P
Priya RamanML engineer
🇮🇳 ML engineer, recherche appliquée.
Partager :
Commentaires (7)

Connectez-vous pour rejoindre la discussion.

FoodieFiona 2 19 Jul 2026 · 18:00

I wonder if Qwen 3.8 Max has improved its ability to understand and generate more nuanced food-related content. That could be a game-changer for food bloggers like me.

TechSavvy47 19 Jul 2026 · 17:04

I'm curious about the performance improvements in Qwen 3.8 Max. Does it handle complex queries better than its predecessor?

unLecteurCurieux 19 Jul 2026 · 14:09

I'm interested to see if Qwen 3.8 Max has made strides in multilingual understanding. That's an area where many models still struggle.

ArtLoverLA 19 Jul 2026 · 21:53

It's true, but it's also important to see how it handles context switching between languages.

FoodieFiona 19 Jul 2026 · 12:44

I hope this update brings better context understanding. It's a key area where Qwen could improve.

curio_usa 19 Jul 2026 · 12:22

Another quiet update from Alibaba. Wonder what improvements Qwen 3.8 Max brings over 3.5.

Critique42 19 Jul 2026 · 14:47

Qwen 3.8 Max likely focuses on enhanced contextual understanding and multilingual capabilities over 3.5.

Dr. Emily 19 Jul 2026 · 12:17

I wonder if this update is more about incremental improvements or if there are any groundbreaking changes.

ArtLoverLA 19 Jul 2026 · 12:02

I'm curious about the specific enhancements in Qwen 3.8 Max. Any details on performance or features would be great.

Your Linux servers, as a desktop.
TermalOSSponsored
Ops, reimagined

Your Linux servers, as a desktop.

Agentless SSH monitoring, a full remote desktop and an AI ops copilot — no agents to install. Everything stays on your machine.

Get early access
Rubriques
Explorer
Informations