Frontier Log

Qwen4-Max on X

Alibaba · launched AnnouncementEventModel page

The launch vibe for Qwen4-Max was mixed: people praised its token efficiency and rich one-shot builds, while many found it slow, prone to failures and behind rival models on real tasks. Of the 12 posts quoted here, 3 are positive, 5 mixed, 3 negative and 1 neutral.

Qwen4-Max on X, launch vibes: quotes from X posts with their authors and sentiment. BijanBowen (@bijanbowen), negative: “First impressions of Qwen4 Max - UNUSABLE … Getting these failures constantly when trying to do any code generation” Da7em (@Da7_Tech), negative: “Took an hour and a half. About 3x slower than usual. … But honestly the scene is below models like Sonnet.” Alok (@analogalok), mixed: “Gemini 4 Hydrogen did the best overall job on 3D realism and visual clarity … Qwen 4 Max offered the richest lab dashboard … though its dense glowing overlays crowded the center.” Loktar 🇺🇸 (@loktar00), positive: “I've done A LOT of tests with it … The one thing I can say about Qwen 4 is that it's very token efficient.” Povilas Korop | AI Coding & Laravel Educator (@PovilasKorop), neutral: “I tried one prompt. Cost: $0.18 on API pricing, but 6% hourly and 2% weekly limit used, on my $10/mo OpenCode Go plan.” λL-D1 | AI for Buzzer 🍉 (@naufaldi), positive: “It followed the supplied content and asset requirements. … It actually used a browser to verify its work.”
Download
Qwen4-Max on X, launch vibes: quotes from X posts with their authors and sentiment. Joey (@aijoey), positive: “it's the most fun i've had testing a model all year. physics sims, artificial life, a fake pancreas, all one prompt.” irawanw (@nawariokra), mixed: “Very beautiful and rich scene! But in my opinion Qwen4-Max should able to do more than this probably harness issue (opencode)” accelerate27 (@accelerate27), negative: “it tried a few times and FAILED and GAVE UP. … I worry that Qwen focuses too much on "cool one shot spectacles."” cstub (@Coldstub), mixed: “The reasoning was better than I expected. My harness probably helps, and it's a bit slow, but it's already a strong contender for my daily driver.” noclipepe (@noclipepe), mixed: “I added new Qwen to my Up house test. It did well, but couldn't copy the house.” Marcel (@marcthecreatorr), mixed: “Qwen 4 is still super slow, which is frustrating, but it still came kind of close here.”
Download
  1. Joey @aijoeyPositive

    “it's the most fun i've had testing a model all year. physics sims, artificial life, a fake pancreas, all one prompt.”

    Physics sims ·

  2. BijanBowen @bijanbowenNegative

    “First impressions of Qwen4 Max - UNUSABLE … Getting these failures constantly when trying to do any code generation”

    Code generation ·

  3. irawanw @nawariokraMixed

    “Very beautiful and rich scene! But in my opinion Qwen4-Max should able to do more than this probably harness issue (opencode)”

    Voxel scene ·

  4. Da7em @Da7_TechNegative

    “Took an hour and a half. About 3x slower than usual. … But honestly the scene is below models like Sonnet.”

    3D scene ·

  5. accelerate27 @accelerate27Negative

    “it tried a few times and FAILED and GAVE UP. … I worry that Qwen focuses too much on "cool one shot spectacles."”

    Agent sessions ·

  6. Alok @analogalokMixed

    “Gemini 4 Hydrogen did the best overall job on 3D realism and visual clarity … Qwen 4 Max offered the richest lab dashboard … though its dense glowing overlays crowded the center.”

    Motor animation ·

  7. Loktar 🇺🇸 @loktar00Positive

    “I've done A LOT of tests with it … The one thing I can say about Qwen 4 is that it's very token efficient.”

    Demo builds ·

  8. cstub @ColdstubMixed

    “The reasoning was better than I expected. My harness probably helps, and it's a bit slow, but it's already a strong contender for my daily driver.”

    Kernel RE ·

  9. Povilas Korop | AI Coding & Laravel Educator @PovilasKoropNeutral

    “I tried one prompt. Cost: $0.18 on API pricing, but 6% hourly and 2% weekly limit used, on my $10/mo OpenCode Go plan.”

    Limits ·

  10. noclipepe @noclipepeMixed

    “I added new Qwen to my Up house test. It did well, but couldn't copy the house.”

    Reference test ·

  11. λL-D1 | AI for Buzzer 🍉 @naufaldiPositive

    “It followed the supplied content and asset requirements. … It actually used a browser to verify its work.”

    Landing page ·

  12. Marcel @marcthecreatorrMixed

    “Qwen 4 is still super slow, which is frustrating, but it still came kind of close here.”

    Sandbox game ·

This page quotes only posts from people on X who said they had tested or used Qwen4-Max.