Qwen 3.7‑Max vs Claude 4.7: 7 In‑Depth Tests Reveal a Smooth, Powerful Model
The author evaluates Alibaba’s newly released Qwen 3.7‑Max across seven rigorous tasks—including reading comprehension, HTML fireworks generation, 3D particle visualizations, PDF‑to‑PPT conversion, Excel data analysis, GitHub trending scraping, and complex video generation—showing it often surpasses GPT‑5.5‑level models and rivals Claude 4.7, especially in long‑duration agent tasks.
Model Overview
Qwen 3.7‑Max (preview) ranks ahead of Kimi‑K2.6, DeepSeek‑v4‑pro and GLM‑5.1 in the Arena global blind‑test leaderboard and is close to GPT, Claude and Gemini among domestic models.
It introduces a new agent architecture that improves programming and reasoning capabilities. On a new chip platform the model performs more than 1,000 tool calls and achieves a ten‑fold inference speed increase compared with its predecessor.
Test 1 – Reading Comprehension, SVG Generation, Aesthetics
The benchmark includes four background figures that many flagship models fail to recognize. Qwen 3.7‑Max handles them correctly and outperforms GPT‑5.5‑level models.
GPT‑5.5 output (both tests forbid any Skills):
Test 2 – Pure HTML Fireworks Effect
> Please write a single‑file dynamic webpage using HTML5, CSS3, and pure JavaScript (Canvas) that creates a dazzling fireworks display.
> 1. Visuals: multiple firework shapes (classic sphere, comet tail, heart), colors generated randomly via HSL, with glowing effects; dark night sky with sparse stars.
> 2. Physics: each particle affected by gravity and air resistance, following realistic parabolic trajectories, with brightness decay and flicker before disappearing.
> 3. Interaction: automatic random launches from the bottom; on any user click/touch, fire a firework at that coordinate.
> 4. Performance: use requestAnimationFrame for smooth animation; bundle all HTML, CSS, and JS into a single index.html file.The model not only follows the prompt but also proactively decides shapes, color schemes, physics and interaction details, demonstrating extraordinary capability.
Test 3 – 900‑Line HTML 3D Particle Galaxy
Prompt: Create a 3D particle galaxy with rotating nebula and dynamic lighting.
Key features generated by the model:
🌀 Spiral‑Arm Structure : five mathematically generated arms with varying rotation speeds.
💫 100k+ Particles : core, arms, halo, dust and nebula layers.
🎨 Dynamic Lighting : five colored point lights moving along trajectories, colors gradually shifting.
🔮 Volumetric Nebula : large semi‑transparent particles simulating gaseous nebula.
✨ Bloom Post‑Processing : Unreal‑style bloom to enhance glow.
🖱️ Interactive Control : mouse drag to rotate, scroll to zoom, auto‑rotate.
🔄 Three Presets : one‑click switch between spiral, elliptical and nebula modes.
🌠 Star‑Trail Effect : optional motion‑trail effect.
⭐ Flicker Animation : each particle independently phases its brightness.
Test 4 – PDF Parsing and PPT Generation
Download .../The Founder’s Playbook: Building an AI‑Native Startup.pdf
Read the full content and, using the html‑ppt‑skill with a tech‑sharing theme, summarize it into a beautiful PPT.The task requires reading a 36‑page PDF; processing time is noticeable.
Test 5 – Excel Processing, Python Data Analysis, Beautiful HTML Report
@2 Excel files contain my public account’s recent 30‑day data. Analyze them, write a richly illustrated data‑analysis report in HTML, focusing on which topics drive reads, shares, and follower growth, and give conclusive recommendations.The model used Pandas to parse the two XLS files, clarified header structures, transformed unstructured Excel data into structured Python data, performed classification and aggregation, and rendered the report with Chat.js, producing a visually appealing HTML page.
Test 6 – Data Crawling, Web Design, Automated Deployment
Fetch the daily top‑10 GitHub trending repositories by star growth, update them periodically, design a webpage, and deploy it online.Because the request was vague, the model invoked its Planning skill, proposed solutions for ambiguous points, and let the user choose. After authorizing a GitHub token, the model pushed the code to a repository, completed deployment, and supplied a step‑by‑step verification workflow.
Test 7 – Ultra‑Long Skills Workflow Execution
A custom “one‑sentence video generation” skill, previously only runnable with Opus or Sonnet 4.7, executed smoothly on Qwen 3.7‑Max.
Signed-in readers can open the original source through BestHub's protected redirect.
This article has been distilled and summarized from source material, then republished for learning and reference. If you believe it infringes your rights, please contactand we will review it promptly.
Old Zhang's AI Learning
AI practitioner specializing in large-model evaluation and on-premise deployment, agents, AI programming, Vibe Coding, general AI, and broader tech trends, with daily original technical articles.
How this landed with the community
Was this worth your time?
0 Comments
Thoughtful readers leave field notes, pushback, and hard-won operational detail here.
