How to build your Qwen 3.8 Max AI system
1. Add Qwen as your main AI brain. 2. Connect Hermes or Claude Code. 3. dd Obsidian memory. 4. Create custom workflows. 5.
Using DeepSeek V4 Flash for this will be cr*zy.
It costs less than Claude Opus 5, Kimi K3, GPT-5.6 Sol, Claude Fable 5, and Qwen 3.8 Max.
Qwen 3.8 Max is one of Alibaba’s biggest AI bets yet.
The numbers are difficult to ignore. → 2.4T total parameters. → Around 95B active per response. → More than 10 days of autonomous coding.
DeepSeek V4 Flash recreated this for far less than Claude Opus 5, Kimi K3, GPT-5.6 Sol, Claude Fable 5, and Qwen 3.8 Max.
That’s what makes its price-to-performance ratio so impressive.
Qwen 3.8 Max can reportedly code for more than 10 days without stopping.
Not 10 hours. 10 days. Alibaba is trying to turn AI from a tool you constantly prompt into a worker that finishes the entire project.
DeepSeek V4 Flash set to increase price.
Will Qwen 3.8 Max, Claude Opus 5, Kimi K3, GPT-5.6 Sol, and Claude Fable 5 match it.
Qwen 3.8 max launch day: the open chinese model where only one AI on earth consistently beats it
This dropped an hour ago. I've been testing it for weeks in preview. Check the Arena leaderboards: → Front-end code: #4, above Fable 5 → Text Arena:…
Qwen 3.8 max VS fable 5: i built 50 tasks on both and the verdict surprised me
The free model won most of the builds. The paid model kept my daily-driver spot anyway.
Qwen 3.8 max free access: the $0 model that matched fable 5 in a side-by-side build test
I ran the same build on both models. Fable 5's version: buggy graphics, movement broken, felt linear.
This AI SEO system ranked #1 on Google in 7 hours, above the AI overviews, using free Qwen 3.8 Max.
The article covered news that dropped the same day. Fully automated. Zero logins to the website.
Qwen 3.8 Max just became the brain you can plug into Hermes with your coding plan, something Fable 5 can't do without the API.
Hermes is the hands. You swap the brain. And this 2.4 trillion parameter brain learned a brand-new skill in 2 minutes flat. Here's what got built
This Agent OS uses one free memory system to train brand-new AI models the second they drop.
Qwen 3.8 Max came out. It got plugged in and knew the whole business instantly. The secret is Obsidian, a free open-source memory vault.
KAT Coder V2.5 Dev (Q4_K_M) - 162 tokens/sec decode - Single RTX 4090 (24GB VRAM) - 8500+ tokens/sec…
KAT Coder V2.5 Dev (Q4_K_M) - 162 tokens/sec decode - Single RTX 4090 (24GB VRAM) - 8500+ tokens/sec prefill - 250k context - almost fully in VRAM -…
Qwen 3.8 Max can now be the brain of your entire agent operating system.
A 2.4 trillion parameter giant plugged into your chat, memory, tickets, and builds. 99% of your competitors don't have this setup.
Qwen 3.8 Max just beat most of 500 human teams in a live contest with a deadline.
And real people ranked it #2 in the world for vision without knowing which model they were rating. Alibaba's new model isn't winning on hype.
Alibaba’s Qwen 3.8 Max worked alone for roughly 125 hours.
It rebuilt a machine-learning experiment, wrote thousands of lines of code and produced improvements that reportedly beat the original method.
Qwen 3.8 Max is an AI with 2.4 trillion parameters that worked alone for 16 days straight.
A year ago, AI lost focus after 30 minutes. This one built a real tool from scratch. Zero humans stepping in.
Alibaba's new Qwen 3.8 Max coded alone for 16 days and shipped finished software.
265 commits. 127 pull requests. Zero human input. It assigned itself the work, wrote the code, ran the tests, and improved on repeat.
Qwen 3.8 Max is a new AI that codes for 10 days straight without you touching it.
Not 10 minutes. Not 10 hours. 10 days. It starts with an empty folder and ends with finished software.
Meta's new Muse Spark 1.2 lost almost every coding test but it wins in one surprising place.
I ran it against GPT 5.6 Soul, Fable 5, and Qwen 3.8 on real builds. The results: → Games came out laggy, buggy, and basic → Qwen 3.8 crushed it on…