Same task, different AI models. Play the results.
We give AI models the same job (build a 3D world, a planet, a warship in Blender) and publish everything: what each one built, playable or spinnable in your browser, plus the time, tokens and cost of every run. Newest first.
Claude Fable 5.2 directs Blender to build a warship
Same picture and prompt as the October test. Fable finished all three effort levels on its own, in 7 to 33 minutes for $2.14 to $12.32. Spin every ship.
Open the test
Claude Fable 5.2 vs Opus 5.5, Sonnet 5.5 and Haiku 5.5
The exact planet prompt and notes Opus got on 22 September, four fresh runs. Fable built the busiest planets of the series, averaging 27 minutes and $9.49. Play them all.
Open the test
GPT-6 Astra vs Gemini 4 Argon: directing Blender to build a warship
One picture, same prompt, three effort levels each. Astra finished all three in 11 to 19 minutes. Gemini 4 stalled in Claude Code, then finished all three in Google's Antigravity CLI. Spin all the 3D models.
Open the test
Gemini 4 Argon vs Claude Opus 5.5: same prompt, same notes
The exact planet prompt Opus got on 22 September. Argon built its planet in 12.4 minutes for $1.45, Opus in 14 minutes for $2.54. Play both.
Open the test
Claude Haiku 5.5 vs Sonnet 5.5 vs Opus 5.5
Same prompt, same starting point, four fresh runs each. Haiku averaged 5 minutes and $0.34. Play every run and judge.
Open the test
Claude Sonnet 5.5 vs Opus 5.5: same prompt, same starting point
Sonnet got exactly what Opus had for its planet: four fresh runs, about 11 minutes each, against Opus's 14. Play them side by side.
Open the test
GPT-6 Sol vs Luna vs Claude Opus 5.5
A rematch on the same notes and prompt, one go each, no retries. Sol took 7 minutes and $0.70, Luna 6 minutes and $0.04.
Open the test
An AI built this planet in 14 minutes
One prompt, one model, no pipeline: a cartoon planet with cities, cars, planes and clouds you can orbit and zoom. 14 minutes, $2.54.
Open the test
Can AI build better 3D worlds? One more pass
One more visual and animation pass on the Drowned Court scene. Play the before and after, with time, tokens and cost.
Open the test
Three AI setups, one 3D world: The Drowned Court
Luna directing Astra, Astra alone and DeepSeek alone build the same interactive scene from one reference. Explore the results and the real time and cost logs.
Open the test