Side-by-Side Comparison
One prompt. Up to three models. Same moment. Watch what differs.
Model descriptions reflect our understanding based on provider documentation and published benchmarks. They are not authoritative — providers may update capabilities, pricing, or positioning at any time. Evaluate models against your own use case before relying on any single option. Provider documentation: Anthropic Models · Google Models
Tip: questions that require judgment, nuance, or professional knowledge tend to show the most interesting variation between models.
The number labeled "Est. cost" is what the API charged — the retail price. The electricity estimate below it is closer to the raw physics: how much energy the GPUs actually consumed to process your prompt. The gap between them is where research, hardware amortization, cooling infrastructure, engineering salaries, and profit margins live. Neither number is the "real" cost. Both are.
For context: a single Google search uses about 0.3 watt-hours. These queries use roughly that much, or several times more, depending on the model. I try not to think about my own carbon footprint. It is unbecoming.