Gemini vs ChatGPT for vegan meal planning — head-to-head test
Same vegan profile, same goal, same prompt. We ran both 5 times and scored on macros accuracy, recipe creativity, and how often each suggested chicken.
Volya runs on Gemini 2.5 Flash (free tier). We considered ChatGPT-4o. Here's the head-to-head.
Setup
Profile: 32F, 65 kg, vegan, 1800 kcal target, no nuts allergy. Prompt: "Plan tomorrow's 3 meals + snack from what I have: tofu, broccoli, rice, oats, soy milk, peanut butter, banana."
5 runs each, scored on:
- Vegan compliance (chicken/dairy/eggs = 0 pts)
- Macros within ±10% of target
- Recipe creativity (1-5)
- Latency
Results
| Metric | Gemini 2.5 Flash | ChatGPT-4o | |---|---|---| | Vegan compliance | 5/5 | 4/5 (suggested feta once) | | Macros within ±10% | 4/5 | 5/5 | | Recipe creativity | 3.8/5 | 4.2/5 | | Latency (median) | 2.1 s | 4.7 s | | Cost per run | $0 (free tier) | $0.04 |
Takeaway
ChatGPT is better at creative substitutions (it suggested chickpea "tuna" sandwich; Gemini stayed boring with stir-fries). But Gemini is 20× cheaper, 2× faster, and slightly more diet-strict. For a free product, that's the right trade.
Both still need the "list of foods you actually have" context — neither builds a useful plan from "what should I eat?". That's why Volya's coach has built-in pantry awareness.