Output Loop Bench · Web promo (HTML)

Electrician promo (6 s, portrait 9:16)

Why we chose this prompt: Portrait is how trades content is watched on phones (reels and stories). It tests whether a model can design for a tall frame in a tight 6 seconds.

ModelMakerTimeTokens
claude-opus-5-5 ★ Output Loop's pickAnthropic218.9s112,275
muse-spark-1.3-mediumMeta21.5s25,559
composer-2.5Cursor23.5s24,105
nemotron-3-nano-4bNVIDIA34.9s3,760
MiniMax-M2.5MiniMax42.1s18,945
gemini-3.8-flashGoogle53.6s21,903
kimi-k3-maxMoonshot (Kimi)60.2s29,783
gemini-4-argon-mediumGoogle65.7s23,152
bonsai-27bPrism ML74.1s4,318
gpt-6-astraOpenAI76.7s18,977
deepseek-v4-proDeepSeek86.2s28,609
gpt-6.1-solOpenAI91.4s18,577
gemma-4-26b-a4b-qatGoogle92.6s10,999
claude-fable-5-1Anthropic107.3s100,143
claude-haiku-5-5Anthropic137.7s115,849
gemini-3.1-proGoogle140.2s22,139
qwen3.8-flashAlibaba (Qwen)191.4s28,722
kimi-k2.7-codeMoonshot (Kimi)196.2s76,461
kimi-k3Moonshot (Kimi)256.5s17,814
muse-glimmerMeta260.4s8,312
gpt-6-lunaOpenAI269.9s28,426
muse-glimmer-30bMeta273.9s19,697
nemotron-3.5-lightning-30b-a3bNVIDIA305.3s31,496
deepseek-v4.1-flashDeepSeek374.2s52,592
grok-4.7-build-fastxAI383.0s64,386
claude-sonnet-5-5Anthropic390.7s140,603
qwen3.8-maxAlibaba (Qwen)418.9s33,685
glm-5.3Z.ai555.2s61,639
grok-4.7xAI571.8s59,841
qwen3.7-plusAlibaba (Qwen)1074.7s44,087
qwen3.8-27bAlibaba (Qwen)1200.5s37,169

Open this task in the interactive Bench