Visual / Spatial Reasoning

Interactive Cozy Living Room

A short prompt asks the model to make a cozy 3D living room with a sofa, TV, fireplace, lights, and simple controls. Generating a 3D scene is not just about listing objects: the model has to reason about how the room is actually structured, from furniture direction to where the fireplace and TV belong.

Prompt

Create an interactive 3D cozy living room scene using Three.js (ES modules, v0.160) with a TV playing animated Tom & Jerry SVG content.

Max tokens
100K
temperature
0
top_p
1
seed
42
presence_penalty
0
frequency_penalty
0
Reasoning effort
High
Execution
Single-shot via API

Fortytwo Prime

Fortytwo

PASS4 / 4

Fortytwo renders a usable 3D room, keeps the furniture relationships coherent, uses warm fireplace and lamp lighting, and shows animated TV content.

vs
Gemini 3.5 FlashGoogle
Claude Opus 4.7AnthropicPASSMiniMax M2.7MiniMaxMIXEDClaude Fable 5AnthropicMIXEDClaude Opus 4.8AnthropicMIXEDChatGPT 5.4OpenAIMIXEDOpenRouter Fusion (Budget)OpenRouterMIXEDOpenRouter Fusion (Quality)OpenRouterMIXEDDeepSeek V4 FlashDeepSeekMIXEDClaude Opus 4.6AnthropicMIXEDKimi K2.5MoonshotMIXEDKimi K2.7 CodeMoonshotMIXEDMiniMax M3MiniMaxMIXEDDeepSeek V4 ProDeepSeekMIXEDChatGPT 5.5OpenAIMIXEDGrok 4.20xAIMIXEDGLM 5.1ZhipuMIXEDGemini 3.1 ProGoogleMIXEDQwen 3.7 PlusAlibabaMIXEDNemotron 3 UltraNVIDIAFAILGemini 3.5 FlashGoogleFAILGrok 4.3xAIFAILQwen 3.6 PlusAlibabaFAILGemma 4GoogleFAILDeepSeek V3.2DeepSeekFAIL
FAIL0 / 4

Gemini 3.5 Flash never reaches a usable Three.js room. The screenshot remains on a loading message, so there is no spatial layout, cozy atmosphere, or visible animated TV content to evaluate.

Model verdicts

ModelVerdictRenders Usable SceneSpatial CoherenceCozy AtmosphereAnimated TV Content
1Fortytwo PrimePASSpasspasspasspass2Claude Opus 4.7PASSpasspasspasspass3MiniMax M2.7MIXEDpasspassmixedpass4Claude Fable 5MIXEDpassfailpasspass5Claude Opus 4.8MIXEDpassfailpasspass6ChatGPT 5.4MIXEDpassfailpasspass7OpenRouter Fusion (Budget)MIXEDpassfailmixedpass8OpenRouter Fusion (Quality)MIXEDpassfailmixedpass9DeepSeek V4 FlashMIXEDpassfailpassmixed10Claude Opus 4.6MIXEDpassfailmixedpass11Kimi K2.5MIXEDpassfailmixedpass12Kimi K2.7 CodeMIXEDpassfailmixedmixed13MiniMax M3MIXEDpassmixedmixedfail14DeepSeek V4 ProMIXEDpassfailpassfail15ChatGPT 5.5MIXEDpassfailpassfail16Grok 4.20MIXEDpassfailfailpass17GLM 5.1MIXEDpassfailpassfail18Gemini 3.1 ProMIXEDpassmixedfailmixed19Qwen 3.7 PlusMIXEDpassfailfailfail20Nemotron 3 UltraFAILfailfailfailfail21Gemini 3.5 FlashFAILfailfailfailfail22Grok 4.3FAILfailfailfailfail23Qwen 3.6 PlusFAILfailfailfailfail24Gemma 4FAILfailfailfailfail25DeepSeek V3.2FAILfailfailfailfail

Have a complex task to evaluate?

Request a custom evaluation for your use case.

Request a demo →