Tetris clone
Also searched as: ai tetris clone coding test
Games4 posts on aivs.tv
The Tetris clone test, run by 5 models across 4 posts on X. Watch each result, then compare what the posts report below.
By model
4 posts3 solo · 1 side-by-sideNot a controlled comparison: prompts, tools and time differ.
Results by model
| Model | Post | Time | Cost | Tokens | Author’s verdict | Original |
|---|---|---|---|---|---|---|
| Claude Opus 5.5 | Hyper-realistic RDR2-style Tetris@SimonasLTU1 | 30 min | — | — | — | X post |
| Claude Opus 5.5 | Self-playing Tetris, 3 Claude models@ka1manov | — | — | — | — | X post |
| Claude Haiku 4.5 | Self-playing Tetris, 3 Claude models@ka1manov | — | — | — | — | X post |
| Claude Sonnet 5.5 | Self-playing Tetris, 3 Claude models@ka1manov | — | — | — | — | X post |
| GLM 5.3 Flash | Tetris on a 1926 newspaper@HelionGlory | — | — | — | — | X post |
| Kimi K2 | Tetris with Kimi K2@cybourgeoisie | — | — | — | Kimi K2 can one-shot coding exercises near the quality of Sonnet 4 | X post |
Only numbers stated in each post. “—” means the post doesn’t say.
Summary
4 posts on X run this test with Claude Opus 5.5, Claude Sonnet 5.5, Claude Haiku 4.5, GLM 5.3 Flash and Kimi K2. @SimonasLTU1 reports per-model numbers (see the table). According to @cybourgeoisie, Kimi K2 can one-shot coding exercises near the quality of Sonnet 4. (The author’s view.)
Who won where
1- Tetris with Kimi K2Kimi K2 can one-shot coding exercises near the quality of Sonnet 4.— the author’s view, @cybourgeoisie (paraphrased)
About this test
Many Tetris videos show a model playing the game rather than building it; those are a different task type (agent play) and are kept apart.
Common wordings
No fixed wording; posts use variations on the same idea.
Create a hyper-realistic Tetris game in the style of RDR2, all in a single HTML file using HTML/CSS/Three.js. — as shown by @SimonasLTU1
Origin: Community prompt (no fixed wording)