← Blog
DESIGN

GLM 5.2 vs DeepSeek V4 Pro vs Kimi K2.7 Code vs MiniMax M3: judged on UX, not just looks

Four open models built the same neon Snake game with Command Code /design, judged specifically on how it feels to play rather than how it screenshots.

Naymur Rahman
Naymur Rahman
3 min read
Jul 16, 2026
Naymur RahmanVerified account
@naymur_dev

GLM 5.2 vs DeepSeek V4 Pro vs Kimi K2.7 Code vs MiniMax M3 I tried creating the same games with each of them. I wanted to see how good they are at UX (not just UI)....

5:17 PM·Jul 16, 2026

Four models, one prompt, one shot each with /design in Command Code - this round deliberately scored on UX over UI.

The prompt

Create a classic Snake game as a single HTML file using Canvas. Grid-based movement, arrow keys or on-screen touch buttons to steer, snake grows as it eats food, game ends on wall or self collision. Neon glow visual style, score display, high score, and a start screen plus game over screen with restart. Make it feel smooth and satisfying to play on both desktop and mobile. Output only the full code.

What we found

GLM 5.2 vs DeepSeek V4 Pro vs Kimi K2.7 Code vs MiniMax M3

I tried creating the same game with each of them. I wanted to see how good they are at UX (not just UI).

For the test, I used the /design from Command Code.

All the models felt similar when I played, but DS4...

The deliberate framing here is worth calling out: most one-shot comparisons in this series (and elsewhere) get judged by screenshot first - does it look polished. This one was explicitly judged by playing it - does moving the snake, turning at speed, and eating food feel right. On a genre as simple as Snake, visual polish is cheap to fake and game feel is not; four models converging on "similar when played" is itself a data point about how far open models have closed the gap on basic game feel, not just visual output.

This is a different four-model field than this repo's other Snake comparisons - Kimi K2.7 Code and MiniMax M3 don't appear in the Muse Spark 1.1 / Fable 5 / Grok 4.5 / GPT-5.6 Sol Snake benchmark elsewhere in this series, so treat the two as complementary fields rather than a single ranking.

Try this yourself

1npm i -g command-code
1/model glm-5.2 2/design Create a classic Snake game as a single HTML file using Canvas. Grid-based movement, arrow keys or on-screen touch buttons to steer, snake grows as it eats food, game ends on wall or self collision. Neon glow visual style, score display, high score, and a start screen plus game over screen with restart. Make it feel smooth and satisfying to play on both desktop and mobile.

Swap /model to compare against DeepSeek V4 Pro, Kimi K2.7 Code, or MiniMax M3. Full command reference: commandcode.ai/docs/slash-commands/design. More one-shot comparisons: CommandCodeAI/slash-design-showcase.

Naymur Rahman
Naymur Rahman@naymur_dev

Share this article