{"slug": "wojak-bench-3d-which-ai-is-best-at-3d-printing", "title": "Wojak-Bench 3D: Which AI is best at 3D printing?", "summary": "A community-voted comparison called Wojak-Bench 3D ranked Claude Fable 5.2 first with a 69% share (9 wins, 0 ties, 4 losses across 13 votes) over GPT-6 Astra at 31% (4/0/9, 13 votes), scoring only same-prompt pairs as (wins + 0.5 × ties) / votes. The benchmark's operators state the results are community preference rather than a controlled evaluation, that small samples are not reliable rankings, and that votes compare photos of physical prints where available and otherwise renders of the print files. The operators also note unequal revision budgets: pointing runs follow one instruction file with equal revision rounds capped at 3, while the shape-sorter pair predates that protocol, with one figure human-iterated over several rounds and the other made directly by the operator, and other pre-protocol figures were excluded.", "body_md": "# which figure matches the drawing better?\n\nnames are hidden until you vote. one vote per browser per pair.\n\n## leaderboard\n\n| # | model | share | w/t/l | votes | \n|---|---|---|---|---|\n| 1 | Claude Fable 5.2 | 69% | 9/0/4 | 13 | \n| 2 | GPT-6 Astra | 31% | 4/0/9 | 13 | \n\n## how this works\n\nshare = (wins + 0.5 × ties) / votes, same-prompt pairs only. a and b are assigned randomly; skips count for nothing. it is community preference, not a controlled evaluation, and small samples are not reliable rankings.\n\nvotes compare photos of physical prints where available, otherwise renders of the print files.\n\nclearing cookies or using another browser can bypass the one-vote limit.\n\n## unequal revision budgets\n\npointing runs follow one instruction file with equal revision rounds (max 3). the shape-sorter pair predates this: one figure was human-iterated over several rounds, the other made directly by the operator. other pre-protocol figures were excluded.", "url": "https://wpnews.pro/news/wojak-bench-3d-which-ai-is-best-at-3d-printing", "canonical_source": "https://wojak-bench.dev.pion.dev/", "published_at": "2026-11-14 19:33:22+00:00", "updated_at": "2026-11-14 19:53:28.610497+00:00", "lang": "en", "topics": ["ai-research", "generative-ai", "large-language-models"], "entities": ["Wojak-Bench 3D", "Claude Fable 5.2", "GPT-6 Astra"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/wojak-bench-3d-which-ai-is-best-at-3d-printing", "markdown": "https://wpnews.pro/news/wojak-bench-3d-which-ai-is-best-at-3d-printing.md", "text": "https://wpnews.pro/news/wojak-bench-3d-which-ai-is-best-at-3d-printing.txt", "jsonld": "https://wpnews.pro/news/wojak-bench-3d-which-ai-is-best-at-3d-printing.jsonld"}}