Flappy Bird (2D/3D)
Also searched as: 3d flappy bird ai coding test
Games4 posts on aivs.tv
The Flappy Bird (2D/3D) test, run by 6 models across 4 posts on X. Watch each result, then compare what the posts report below.
By model
4 posts2 solo · 2 side-by-sideNot a controlled comparison: prompts, tools and time differ.
Results by model
| Model | Post | Time | Cost | Tokens | Author’s verdict | Original |
|---|---|---|---|---|---|---|
| Claude Fable 5 | One-shot Flappy Bird with scores@AiHubMix | — | $0.55 | — | — | X post |
| Claude Fable 5 | Flappy Bird@bridgebench | — | — | — | — | X post |
| Claude Opus 5 | Flappy Bird, 3 models with costs@noclipepe | — | $0.253 | — | — | X post |
| Claude Sonnet 4 | Flappy Bird in pure HTML/CSS/JS@scaling01 | — | — | — | — | X post |
| DeepSeek V4 Pro 0813 | One-shot Flappy Bird with scores@AiHubMix | — | $0.02 | — | Author’s pick | X post |
| GPT-5.6 Sol | Flappy Bird, 3 models with costs@noclipepe | — | $0.15 | — | — | X post |
| Qwen 3.8 Max | Flappy Bird, 3 models with costs@noclipepe | — | $0.0248 | — | — | X post |
Only numbers stated in each post. “—” means the post doesn’t say.
Summary
4 posts on X run this test with DeepSeek V4 Pro 0813, Claude Fable 5, Qwen 3.8 Max, GPT-5.6 Sol, Claude Opus 5 and Claude Sonnet 4. @AiHubMix, @noclipepe report per-model numbers (see the table). According to @noclipepe, Qwen matched GPT on gameplay and UX/UI for 6x less. (The author’s view.) According to @bridgebench, the best UI they have ever seen from an AI model; no other model was close. (The author’s view.)
Who won where
2- Flappy Bird, 3 models with costsQwen matched GPT on gameplay and UX/UI for 6x less.— the author’s view, @noclipepe (paraphrased)
- Flappy BirdThe best UI they have ever seen from an AI model; no other model was close.— the author’s view, @bridgebench (paraphrased)
About this test
A small game that exposes physics tuning and collision bugs quickly.
Canonical prompt
Build a fully playable 3D Flappy Bird game in a single HTML file.
Origin: @ivanainai