Side-by-side comparison of outputs from multiple language models for the same prompt — model name, output text, and optional latency/cost/token metadata per column. Designed for model evaluation, A/B testing, and capability demonstrations. Original a2ui-catalogue atom.
| Surface | Note |
|---|---|
| googlechat | renders as sequential key-value blocks; side-by-side layout not supported |
| multi-column layout collapses in most email clients; renders as stacked blocks |
| Field | Type | |
|---|---|---|
| prompt | string. The shared input prompt shown above the comparison. | optional |
| models | {'type': 'array', 'description': 'List of {name, output, latency_ms?, cost_usd?, tokens?} model result objects.'} | required |
| show_meta | boolean. Show latency/cost/token row below each output. Default true if any model provides meta fields. | optional |
{
"type": "llm_comparison_table",
"models": [
{
"name": "GPT-4",
"context": "128k"
},
{
"name": "Claude Sonnet",
"context": "200k"
}
]
}{
"identifier": "urn:air:a2uicatalog.ai:atom:llm_comparison_table",
"displayName": "Llm Comparison Table",
"type": "application/vnd.a2ui.atom+json",
"url": "https://a2uicatalog.ai/atoms/llm_comparison_table",
"capabilities": [
"web",
"meet-stage",
"pdf",
"apps-script-web"
],
"description": "Side-by-side comparison of outputs from multiple language models for the same prompt \u2014 model name, output text, and optional latency/cost/token metadata per column. Designed for model evaluation, A/B testing, and capability demonstrations. Original a2ui-catalogue atom.",
"representativeQueries": [
"show a side-by-side multi-model output comparison table",
"side-by-side comparison of outputs from multiple language models for the same prompt \u2014 model name, o",
"render a llm comparison table"
]
}The renderer is open source. Deploy your own instance in 4 commands — you own the URL, no dependency on the demo endpoint.
git clone https://github.com/a2uicatalog/a2ui
cd apps-script-surface/gas-schema-renderer
clasp push && clasp deploy