Output Loop Bench · Workflows

Page manager: build

Why we chose this prompt: Frontier goes one step further than Fast: the workflow also runs the page day to day. Step 1 of 2: the model writes and saves the workflow; the Lab then runs it on a simulated day it never saw.

ModelMakerTimeTokens
kimi-k2.7-codeMoonshot (Kimi)169.4s603,831
claude-sonnet-5-5Anthropic203.5s1,771,958
MiniMax-M2.5MiniMax248.8s455,507
claude-opus-5-5Anthropic258.2s1,902,646
gpt-6.1-solOpenAI263.0s375,261
qwen3.7-plusAlibaba (Qwen)285.0s598,586
claude-fable-5-1Anthropic291.5s1,623,046
gpt-6-astraOpenAI359.5s891,683
deepseek-v4-proDeepSeek538.3s1,692,697
qwen3.8-maxAlibaba (Qwen)1139.3s1,225,848
glm-5.3Z.ai1207.5s581,596
grok-4.7xAI2128.6s4,618,294

Open this task in the interactive Bench