We generated Japanese businesspeople spanning 20s, 40s and 60s across male and female genders in a Tokyo office lobby using the same prompt structure. GPT Image 2 scored highest on average (4.04) and won all six prompts, closely followed by NanoBanana 2 (4.00). FLUX.2 [dev] scored 3.67 but was the cheapest at about $0.012 per image. All three models produced commercially usable results across age and gender variations.
Results
| Model | Avg Score | Wins | Cost / image | Verdict |
|---|---|---|---|---|
| GPT Image 2 | 4.04 | 6 | $0.0800 | Usable across all age/gender groups |
| NanoBanana 2 | 4.00 | 0 | $0.0800 | Consistently usable |
| FLUX.2 [dev] | 3.67 | 0 | $0.0120 | Cheapest, slightly lower score |
Key takeaways
- GPT Image 2 is the most stable for Japanese businesspeople across age and gender, winning every demographic tested.
- NanoBanana 2 matched GPT Image 2 closely with a 4.00 average, making it a strong alternative.
- FLUX.2 [dev] remains the lowest-cost option at about $0.012 per image with acceptable quality.
- Age and gender variations are handled well by all three models; the biggest differentiator is cost and subtle quality.
Verdict
For Japanese businessperson generation across ages and genders, GPT Image 2 is the top choice for quality, NanoBanana 2 is a solid runner-up, and FLUX.2 [dev] is the budget option for high-volume generation.
Note
This is an English summary of the full Japanese benchmark article. Methodology, prompts, all outputs and cost details are documented in the original post.
Read the full article in Japanese →