Output Loop Bench · Artifacts
SVG drawing: AI models compared
A drawing made entirely from code. 30 models tested on the same task.
Pelican riding a bicycle
The famous public test that appears at almost every model launch. Weak models draw blobs or broken bikes; strong ones get the anatomy and the bike's geometry right.
| Model | Maker | Time | Tokens |
|---|---|---|---|
| gemini-4-argon-medium ★ Output Loop's pick | 171.5s | 36,121 | |
| claude-sonnet-5-5 ★ Output Loop's pick | Anthropic | 272.9s | 123,464 |
| muse-spark-1.3-medium | Meta | 29.2s | 28,242 |
| composer-2.5 | Cursor | 30.3s | 25,211 |
| nemotron-3-nano-4b | NVIDIA | 37.4s | 3,900 |
| gemma-4-26b-a4b-qat | 49.3s | 6,044 | |
| kimi-k3-max | Moonshot (Kimi) | 74.1s | 31,835 |
| nemotron-3.5-lightning-30b-a3b | NVIDIA | 86.6s | 21,737 |
| MiniMax-M2.5 | MiniMax | 95.6s | 21,736 |
| gemini-3.8-flash | 106.0s | 29,380 | |
| gpt-6-astra | OpenAI | 117.1s | 19,871 |
| gpt-6.1-sol | OpenAI | 136.6s | 20,640 |
| qwen3.8-flash | Alibaba (Qwen) | 143.2s | 25,489 |
| claude-opus-5-5 | Anthropic | 164.9s | 104,599 |
| bonsai-27b | Prism ML | 185.3s | 10,211 |
| deepseek-v4-pro | DeepSeek | 185.9s | 39,972 |
| muse-glimmer | Meta | 193.6s | 6,214 |
| gemini-3.1-pro | 215.2s | 26,359 | |
| kimi-k2.7-code | Moonshot (Kimi) | 248.7s | 64,815 |
| claude-fable-5-1 | Anthropic | 257.2s | 109,957 |
| gpt-6-luna | OpenAI | 270.9s | 27,548 |
| claude-haiku-5-5 | Anthropic | 281.9s | 146,969 |
| kimi-k3 | Moonshot (Kimi) | 304.4s | 18,766 |
| muse-glimmer-30b | Meta | 330.7s | 18,852 |
| deepseek-v4.1-flash | DeepSeek | 356.1s | 54,270 |
| grok-4.7 | xAI | 356.7s | 50,736 |
| qwen3.7-plus | Alibaba (Qwen) | 362.6s | 37,510 |
| grok-4.7-build-fast | xAI | 404.7s | 77,935 |
| glm-5.3 | Z.ai | 495.5s | 54,657 |
| qwen3.8-max | Alibaba (Qwen) | 2238.4s | 57,301 |