MethodologyAnnotator qualityUpdatesHugging Face436.7k downloadsSuggest a model or benchmarkRequest a private benchmark
Benchmarks
Dimensions
License
Models
Categories
category
source
LLM to Animation v1updated 2026-11-19

Language Models: Animation Benchmark

Rapidatabenchmark@rapidata.ai

Abstract

When a new language model comes out, one of the most common ways to show it off is a three.js animation: a single clip of a scene the model wrote in code, chosen by the person posting it. This benchmark ranks language models on that task under fixed conditions. Every model receives the same 200 prompts, drawn from requests people post on X and type into chatbots and widened with scene descriptions from video benchmarks, and writes one HTML page per prompt that plays a three.js animation. Each page is recorded as a 7-second clip, and human annotators compare the clips of two models for the same prompt, without seeing the prompt or the code, and say which one they would continue watching and which one they like more. The 925,025 responses collected this way are aggregated into ratings with a Bradley–Terry model; the details describe how the prompts were chosen and the clips made, and the methodology covers matchmaking and score estimation.

LeaderboardCompare OutputsMatchupsAnalysisPromptsModel settingsDetails

29 models. Settings sent per model, besides the system prompt (listed under Details) and the prompt. Each model wrote an HTML page, recorded as a 7-second clip.

Model
Endpoint
Settings sent
Output
qwen3.8-max-0902
OpenRouter qwen/qwen3.8-max-0902
reasoning effort high, max tokens 128,000 (thinking included)
MP4
qwen4-max
OpenRouter qwen/qwen4-max
reasoning effort high, max tokens 128,000 (thinking included)
MP4
claude-fable-5.1
OpenRouter anthropic/claude-fable-5.1
reasoning effort high, max tokens 128,000 (thinking included)
MP4
claude-fable-5.2
OpenRouter anthropic/claude-fable-5.2
reasoning effort high, max tokens 128,000 (thinking included)
MP4
claude-haiku-5.5
OpenRouter anthropic/claude-haiku-5.5
reasoning effort high, max tokens 128,000 (thinking included)
MP4
claude-opus-5.5
OpenRouter anthropic/claude-opus-5.5
reasoning effort high, max tokens 128,000 (thinking included)
MP4
claude-sonnet-5.5
OpenRouter anthropic/claude-sonnet-5.5
reasoning effort high, max tokens 128,000 (thinking included)
MP4
deepseek-v4-pro-0813
OpenRouter deepseek/deepseek-v4-pro-0813
reasoning effort high, max tokens 128,000 (thinking included)
MP4
deepseek-v4.1-flash-1017
OpenRouter deepseek/deepseek-v4.1-flash-1017
reasoning effort high, max tokens 128,000 (thinking included)
MP4
gemini-3.8-flash
OpenRouter google/gemini-3.8-flash
reasoning effort high, max tokens 128,000 (thinking included)
MP4
gemini-4-argon
OpenRouter google/gemini-4-argon
reasoning effort high, max tokens 128,000 (thinking included)
MP4
gemini-4-carbon
Vertex AI google/gemini-4-carbon
reasoning effort high, or medium where high gave no usable page; max tokens 128,000 requested, capped by Vertex AI at 65,536
MP4
gemini-4-hydrogen
OpenRouter google/gemini-4-hydrogen
reasoning effort high, max tokens 128,000 (thinking included)
MP4
muse-ember
OpenRouter meta/muse-ember
reasoning effort high, max tokens 128,000 (thinking included)
MP4
muse-spark-open
OpenRouter meta/muse-spark-open
reasoning effort high, max tokens 128,000 (thinking included)
MP4
muse-wonder
OpenRouter meta/muse-wonder
reasoning effort high, max tokens 128,000 (thinking included)
MP4
muse-wonder-full
Clips uploaded directly, not generated through OpenRouter
Not recorded with the benchmark
MP4
minimax-m3.1-flash
OpenRouter minimax/minimax-m3.1-flash
reasoning effort high, max tokens 128,000 (thinking included)
MP4
mistral-large-4-0
OpenRouter mistralai/mistral-large-4-0
reasoning effort high, max tokens 128,000 (thinking included)
MP4
kimi-k3.1-pro
OpenRouter moonshotai/kimi-k3.1-pro
reasoning effort high, max tokens 128,000 (thinking included)
MP4
gpt-6-astra
OpenRouter openai/gpt-6-astra
reasoning effort high, max tokens 128,000 (thinking included)
MP4
gpt-6-luna
OpenRouter openai/gpt-6-luna
reasoning effort high, max tokens 128,000 (thinking included)
MP4
gpt-6.1-sol
OpenRouter openai/gpt-6.1-sol
reasoning effort high, max tokens 128,000 (thinking included)
MP4
gpt-6.1-terra
OpenRouter openai/gpt-6.1-terra
reasoning effort high, max tokens 128,000 (thinking included)
MP4
gpt-6.2-astra-pro
OpenRouter openai/gpt-6.2-astra-pro
reasoning effort high, max tokens 128,000 (thinking included)
MP4
gpt-6.2-sol-pro
OpenRouter openai/gpt-6.2-sol-pro
reasoning effort high, max tokens 128,000 (thinking included)
MP4
mimo-v2.6-pro
OpenRouter xiaomi/mimo-v2.6-pro
reasoning effort high, max tokens 128,000 (thinking included)
MP4
glm-5.3
OpenRouter z-ai/glm-5.3
reasoning effort high, max tokens 128,000 (thinking included)
MP4
grok-4.7
OpenRouter x-ai/grok-4.7
reasoning effort high, max tokens 128,000 (thinking included)
MP4
qwen3.8-max-0902 MP4
OpenRouter qwen/qwen3.8-max-0902
reasoning effort high, max tokens 128,000 (thinking included)
qwen4-max MP4
OpenRouter qwen/qwen4-max
reasoning effort high, max tokens 128,000 (thinking included)
claude-fable-5.1 MP4
OpenRouter anthropic/claude-fable-5.1
reasoning effort high, max tokens 128,000 (thinking included)
claude-fable-5.2 MP4
OpenRouter anthropic/claude-fable-5.2
reasoning effort high, max tokens 128,000 (thinking included)
claude-haiku-5.5 MP4
OpenRouter anthropic/claude-haiku-5.5
reasoning effort high, max tokens 128,000 (thinking included)
claude-opus-5.5 MP4
OpenRouter anthropic/claude-opus-5.5
reasoning effort high, max tokens 128,000 (thinking included)
claude-sonnet-5.5 MP4
OpenRouter anthropic/claude-sonnet-5.5
reasoning effort high, max tokens 128,000 (thinking included)
deepseek-v4-pro-0813 MP4
OpenRouter deepseek/deepseek-v4-pro-0813
reasoning effort high, max tokens 128,000 (thinking included)
deepseek-v4.1-flash-1017 MP4
OpenRouter deepseek/deepseek-v4.1-flash-1017
reasoning effort high, max tokens 128,000 (thinking included)
gemini-3.8-flash MP4
OpenRouter google/gemini-3.8-flash
reasoning effort high, max tokens 128,000 (thinking included)
gemini-4-argon MP4
OpenRouter google/gemini-4-argon
reasoning effort high, max tokens 128,000 (thinking included)
gemini-4-carbon MP4
Vertex AI google/gemini-4-carbon
reasoning effort high, or medium where high gave no usable page; max tokens 128,000 requested, capped by Vertex AI at 65,536
gemini-4-hydrogen MP4
OpenRouter google/gemini-4-hydrogen
reasoning effort high, max tokens 128,000 (thinking included)
muse-ember MP4
OpenRouter meta/muse-ember
reasoning effort high, max tokens 128,000 (thinking included)
muse-spark-open MP4
OpenRouter meta/muse-spark-open
reasoning effort high, max tokens 128,000 (thinking included)
muse-wonder MP4
OpenRouter meta/muse-wonder
reasoning effort high, max tokens 128,000 (thinking included)
muse-wonder-full MP4
Clips uploaded directly, not generated through OpenRouter
Not recorded with the benchmark
minimax-m3.1-flash MP4
OpenRouter minimax/minimax-m3.1-flash
reasoning effort high, max tokens 128,000 (thinking included)
mistral-large-4-0 MP4
OpenRouter mistralai/mistral-large-4-0
reasoning effort high, max tokens 128,000 (thinking included)
kimi-k3.1-pro MP4
OpenRouter moonshotai/kimi-k3.1-pro
reasoning effort high, max tokens 128,000 (thinking included)
gpt-6-astra MP4
OpenRouter openai/gpt-6-astra
reasoning effort high, max tokens 128,000 (thinking included)
gpt-6-luna MP4
OpenRouter openai/gpt-6-luna
reasoning effort high, max tokens 128,000 (thinking included)
gpt-6.1-sol MP4
OpenRouter openai/gpt-6.1-sol
reasoning effort high, max tokens 128,000 (thinking included)
gpt-6.1-terra MP4
OpenRouter openai/gpt-6.1-terra
reasoning effort high, max tokens 128,000 (thinking included)
gpt-6.2-astra-pro MP4
OpenRouter openai/gpt-6.2-astra-pro
reasoning effort high, max tokens 128,000 (thinking included)
gpt-6.2-sol-pro MP4
OpenRouter openai/gpt-6.2-sol-pro
reasoning effort high, max tokens 128,000 (thinking included)
mimo-v2.6-pro MP4
OpenRouter xiaomi/mimo-v2.6-pro
reasoning effort high, max tokens 128,000 (thinking included)
glm-5.3 MP4
OpenRouter z-ai/glm-5.3
reasoning effort high, max tokens 128,000 (thinking included)
grok-4.7 MP4
OpenRouter x-ai/grok-4.7
reasoning effort high, max tokens 128,000 (thinking included)

We use crowd intelligence to deliver real, diverse human feedback at scale for model evaluation and post-training.

Benchmarks
Text-to-ImageOutput Diversity (Mode Collapse)Multi-Step Image Editing (Identity Drift)Camera MovementRelative Camera MovementPhysicsPhysicsRealismImage-to-3DSVGLLM-to-3DLLM to AnimationSuggest a model or benchmark
Method
MethodologyData on Hugging Face
Rapidata
rapidata.ai Private benchmarksContactX LinkedIn
Rapidata, benchmark.ai © 2026