Reddit builder compares Grok 4.5, Qwen 3.8 Max, GLM 5.2 and GPT 5.3 Codex on a real Three.js coding challenge
EarlyBuilder_529 · reddit · 2026-07-25
A Reddit builder ran a real-world AI coding battle on the same Three.js prompt and compared Grok 4.5, Qwen 3.8 Max, GLM 5.2, GPT 5.3 Codex, and Claude Sonnet 5 on visual quality, prompt understanding, animation flow, engineering execution, code quality, and UX.
- Prompt: a fully self-contained single-file HTML demo that renders a procedural 3D Earth, then launches a meteor impact, explosive flash, and 300+ debris particles.
- Winner: Grok 4.5, for the smoothest animation and strongest cinematic presentation.
- Runner-up: Qwen 3.8 Max, which the author said was the biggest surprise.
- Also noted: GLM 5.2 had strong particles and visuals; GPT 5.3 Codex was cleanly engineered but ranked lower in this graphics-heavy test.
- The author says the ranking is prompt-specific and plans this as Episode 1 of a broader coding battle series.
More from coding & agent
- Hermes pitch claims persistent memory and skill generation can make agents self-improving — Teknium · 2026-07-25
- Cass aims to search across Claude Code, Codex, Cursor and Gemini CLI sessions — doodlestein · 2026-07-25
- GlobalGPT launches a CLI that plugs skills and MCP into Codex — HeyAmit_ · 2026-07-25
- GPT-5.6 and Claude Opus 5 both built a Flappy Bird clone from one prompt — Arindam_1729 · 2026-07-25
- AI agents that read 13F filings often misread puts as bullish long positions — Capedcrusader1923 · 2026-07-25
- MCP’s real business value may be in trust and permissions, not connectors — Warm-Reaction-456 · 2026-07-25