GPT Image 2.6 Flare 生成速度较 2.5 Flare 提升近一倍
OpenAI 新的 GPT Image 2.6 Flare 把最高质量出图从 55 秒压到 30 秒，文生图和图像编辑两套基准分数也涨了。做批量生图或编辑功能的开发者值得跑一次对比测试，速度翻倍对产品体验差别很大。
Artificial Analysis 的独立基准测试显示，GPT Image 2.6 Flare 在最高质量档的中位生成时间约为 30 秒，而 GPT Image 2.5 Flare 约为 55 秒，速度提升接近一倍。
除速度外，新版本在两项质量基准上也有提升：AA-Image-T2I v2.1（文生图）和 AA-Image-Editing v2.1（图像编辑）的得分均高于 GPT Image 2.5 Flare，说明这次升级在速度和质量两条线上同步改进。
对依赖批量生图或实时编辑工作流的开发者来说，30 秒与 55 秒的差距会直接影响产品响应体验；在同类图像生成 API 的横向比较中，速度翻倍叠加质量得分上涨，使这一版本在选型时更具竞争力。
GPT Image 2.6 Flare generates images almost twice as fast as GPT Image 2.5 Flare
Median generation time in our independent benchmark, at max quality: about 30 seconds for GPT Image 2.6 Flare, about 55 seconds for GPT Image 2.5 Flare.
And it scores higher on both AA-Image-T2I v2.1 and AA-Image-Editing v2.1 too!
This report is categorized under AITOP's "Models" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows and US-China foundation model benchmarks and commercialization comparison. Key entities and technical topics under continuous tracking include GPT Image 2.6 Flare, OpenAI, 文生图, 图像生成.
Source attribution: curated from Artificial Analysis (@ArtificialAnlys).
