opus-5.5VIDEO
10 models tested on the same Blender claymation script prompt
Ten models were given the same prompt to write a Blender script for a 96-frame looping claymation desk scene; Opus 5.5, Sonnet 5.5 and four others passed, while GLM, Composer and Qwen failed.
Wayne Lowry
@wikiwayneArchived 20 days agoOriginally posted 2026.10.08
Wayne Lowry @wikiwayneOriginally posted 2026.10.08
I gave 10 models the same prompt: one Python script that builds a looping claymation desk scene in Blender. A clay mug with steam, a plant with five leaves, and a paper airplane circling the mug. 96 frames, animated on twos. The rules: no edits. I ran every script exactly as written, in Blender 4.5 and 5.1, on my GX10. Passed in 4.5: Opus 5.5 Grok 4.7 Sonnet 5.5 Muse Spark 1.3 Haiku 5.5 GPT-5.6 Sol Failed: GLM 5.3 Flash Composer 2.5 Grok 4.6 Qwen 3.8 Flash Next (local, even after 23 rounds of fixing its own errors)
x.com/wikiwayne/status/2107951058584486055Archived 2026.10.08 10:30