Claude Fable 5.2 Tops AA-3D Leaderboard with 67-Point Elo Lead
Artificial Analysis announced that Claude Fable 5.2 (Max) has taken the top spot on its 3D interactive experience building benchmark AA-3D, with an Elo score of 1271 — 67 points ahead of runner-up Claude Sonnet 5.5 and 132 points above the previous-generation Claude Fable 5.1 (1139 Elo).
Confirmed
- AA-3D's evaluation pipeline is build, check, vote, rank: models run as agents in an offline sandbox, turning text briefs into three.js web pages, and can self-test and iterate using Blender and a browser.
- The brief only specifies what the scene must contain; everything else is up to the model. Examples include a full Starship mission animation (launch to orbit, booster tower catch recovery, ship landing) and a tour of the Taj Mahal.
- In hands-on comparisons, Claude Fable 5.2 showed improved accuracy on game and fantasy-story scenarios, winning same-brief matchups including a dodgeball game with a live scoreboard, an octopus playing the drums, and a dragon breathing fire on a medieval town, with topics also covering world wonders.
- In a same-brief comparison (two disk galaxies inspired by the Tentacle Galaxy colliding and merging, same tools and time limit), the gap between the two generations was stark: 1271 vs. 1139, with clearly different levels of detail in the galaxy collision scene.
Why it matters
- AA-3D tests a model's ability to autonomously complete complex 3D interactive builds as an agent, not single-turn generation; Claude Fable 5.2's wide margin signals a generational leap in long-horizon task planning and fine-grained visual control.
2026-11-10 ~ 2026-11-10 · 6 related posts
Primary sources
- Claude Fable 5.2 Tops AA-3D Benchmark With 1271 Elo, 67 Points Ahead — ArtificialAnlys ·
- How AA-3D Works: Models as Agents in Sandboxes Building 3D Scenes — ArtificialAnlys ·
- Two Generations Compared: Claude Fable 5.2 vs 5.1 on the Same 3D Brief — ArtificialAnlys ·
- [source] Claude Fable 5.2 Tops AA-3D Benchmark With 1271 Elo, 67 Points Ahead — ArtificialAnlys · 2026-11-10
- [source] Two Generations Compared: Claude Fable 5.2 vs 5.1 on the Same 3D Brief — ArtificialAnlys · 2026-11-10
- AA-3D Briefs Explained: From Starship Missions to Taj Mahal Walkthroughs — ArtificialAnlys · 2026-11-10
- Claude Fable 5.2 tops rivals in live comparisons: dodgeball, drumming octopus, fire-breathing dragon — ArtificialAnlys · 2026-11-10
- Claude Fable 5.2 Also Leads in Games and Fantasy Scene Building — ArtificialAnlys · 2026-11-10
- [source] How AA-3D Works: Models as Agents in Sandboxes Building 3D Scenes — ArtificialAnlys · 2026-11-10