Qwen4-Max hands-on: massive gains over Qwen3.8-Max, even in embedded dev

karminski3 · x · 2026-11-20

Bilibili creator karminski-牙医 published a full hands-on review of the Qwen4-Max release. Having tested the model during internal beta, he re-ran the same one-shot ray-tracing prompt after launch and found a dramatic gap — the final version is far better — so he re-did all his tests.

The review covers:

Verdict: huge improvement over Qwen3.8-Max, with especially notable gains in frontend and agent capabilities; backend agentic coding "seems promising" — the video examines whether it can serve as a daily-driver coding model.

Related event: Hands-on Tests Show Qwen4-Max Makes Big Gains in Frontend and Agent Skills(2 posts)→

Original post →

More from Models

Models channel →