MiniMax M2.1 vs Opus 4.5 vs GLM-4.7 built an interactive 3D solar system from scratch.
@ivanfioravanti45K followers
81,820 views
47 Comments
Comments from the original post on X (Twitter)

@ivanfioravanti
Prompt: Build an interactive 3D solar system visualization using Three.js and vanilla JavaScript in a single HTML file. Requirements: - All 8 planets with accurate relative sizes and orbital distances (use logarithmic scale for visibility) - The Sun at the center with a glowing effect (emissive material + point light) - Planets orbit the Sun at different speeds (Mercury fastest, Neptune slowest) - Each planet should have: - Accurate colors/textures (use procedural or solid colors, no external images) - Axial tilt and rotation - Saturn with rings - Earth with a moon orbiting it - Interactive controls: - OrbitControls for camera (zoom, pan, rotate) - Click on a planet to show an info card (name, diameter, distance from sun, orbital period, fun fact) - Time speed slider (0.1x to 10x) to speed up/slow down orbits - Pause/Play button - Starfield background (particle system) - Smooth animations at 60fps - Beautiful UI overlay with: - Title "Solar System Explorer" - Current simulation speed display - Planet selector dropdown to focus camera on a planet - Dark theme, modern aesthetic with subtle glow effects - Responsive design Make it visually stunning.
29 likes

@ivanfioravanti
GLM 4.7 is the second demo in the video.
19 likes

@EAccelerate_42
GLM 4.7 is too good Try the models to “make a self playing chess game”. In my test surprisingly even opus and Gemini made game were doing random movement and only GLM4.7 made game that played exactly with correct chess movements
13 likes

@ivanfioravanti
Grate idea! I will try this one. I'm starting to use GLM-4.7 more now and I love the design capabilities!
3 likes

@KuittinenPetri
GLM-4.7 and even Qwen Coder or Qwen3 does better looking web layouts than most Western models. And you can run those Qwen models on your own computer with 32 GB RAM. Qwen3-VL-30B-A3B performs well on turn a sketch or screenshot into fully working HMTL5 & JS app tests.
10 likes

@ivanfioravanti
What a great suggestion! Use GLM for design only in case of need!
2 likes

@KuittinenPetri
In my experience that GPT-5.1-codex (max) and gpt-5-pro and Claude 4.5 Opus are the best at fixing bugs, but they are not the best in design. 1. Do layout and design using cheap Chinese OSS. 2. Fix the difficult bugs using expensive, slow and smart closed source models.
4 likes

@ctodotnew
gotta run it back
4 likes

@itssidhere
so should i cancel my m2.1 and switch to glm4.7?
3 likes

@ivanfioravanti
I have both. M2.1 is super fast in coding, glm-4.7 is strong in design.
9 likes

@CanopyWave_CW
Speed vs quality is always a tough balance. Cool to see GLM 4.7 nail it.
2 likes

@TermLimitsNow48
now try, create a complex derivative trading system, with compliance , P&L calculations supervisory controls, real-time pricing and iterate over the platform no less than 300 times
2 likes

@tryteraai
MiniMax M2.1 vs Opus 4.5 vs GLM-4.7 built a 3D solar system. Running 3 Claude Codes in parallel. 🥇 GLM-4.7 – design skills next level. 🥈 Opus 4.5 – slowest. 🥉 M2.1 – fastest.
2 likes

@ivanfioravanti
Can't wait to enter in 2026 and deep dive on AI stuff!
2 likes

@skillcreatorai
m2.1 fastest but still bronze. speed doesn't win races apparently
2 likes

@dappsidesanchez
GLM-4.7 is on another level, the designs are immaculate bro! 100%
1 like

@thekennwakanma
have you tried glm and m2.1 on real tough code debugs? secondlt would you say glm is almost as good as Gemini3 flash for ui?
1 like

@ivanfioravanti
Yes for m2.1, using their API, not local inference and it really rocks. Model tends to do much more than asked for, you have to tame it, but it's solid and fast. GLM great for design, I was really impressed. Gemini is still a step above anything else on this front IMO.

@thekennwakanma
Okay thank you for the unbiased review chief. Connect?

@finityx
Put it against me
1 like


















