Gemini 1.5 Pro JSON Mode vs Tablesmith
Yesterday, I released Tablesmith 1.1.2, featuring a new context-aware autofill functionality. To showcase this new feature, I recorded a video demonstrating recipe generation.
Interestingly, Google’s # Gemini 1.5 Pro also added a JSON mode around the same time, which utilizes a similar recipe example as well. Let’s see how they compare!
Tablesmith
Token used: input-35430, output-7141
Estimated cost: $0.028
Gemini 1.5 Pro
Result (converted to CSV for readability)
Token used: input-300, output-6126
Estimated cost: $0.124 (24 recipes)
Note: I requested 30 recipes in both the prompt and JSON schema, but Gemini Pro only returned 24.
Prompt:
Create a list of random popular Thai or Japanese dishes found in American restaurants.
One Japanese recipe followed by one Thai recipe, alternating.
Give me 30 recipes and using this Json schema: {
"$schema": "http://json-schema.org/draft-07/schema#",
"title": "Generated schema for Root",
"type": "array",
"minItems": 30,
"maxItems": 30,
"items": {
"type": "object",
"properties": {
"name": {
"type": "string"
},
"ingredients": {
"type": "array",
"items": {
"type": "string"
}
},
"steps": {
"type": "array",
"items": {
"type": "string"
}
},
"tags": {
"type": "array",
"items": {
"type": "string"
}
},
"nutrition": {
"type": "object",
"properties": {},
"required": []
}
},
"required": [
"name",
"ingredients",
"steps",
"tags",
"nutrition"
]
}
}
Each step in steps field mention the ingredients you'll be using by placing the ingredient name following a corresponding number. For example: "Brown the ground beef [1] and onion [2] in skillet."
Tags field contains 5 tags.
Nutrition field include CALORIES,FAT,PROTEIN and CARBS
Gemini Pro 1.5 delivers highly accurate JSON output. It provides well-formatted ingredients and steps in clear JSON arrays, ensuring consistent formatting across all recipes.