WHAT NEW IN X10.31 SAT

X 上的 AI 合集档案 · 22 个合集 · 61,986 条

gpt-6IMAGE

GPT-6 Astra 与 Image 2.5 同周发布,Codex 内置生图被夸

作者对比 Codex、MiniMax、Grok 的内置生图能力,称 Codex 最强,并用 Image 2.5 的 Sketch 模式把涂鸦改成《恶魔之魂》Boss 战画面。

ashen
@ashen_one2026.09.14收录原推发布于 2026.09.09
GPT-6 Astra 与 Image 2.5 同周发布,Codex 内置生图被夸(第 1 张,共 2 张)
GPT-6 Astra 与 Image 2.5 同周发布,Codex 内置生图被夸(第 2 张,共 2 张)
ashen @ashen_one原推发布于 2026.09.09
在 X 上打开原推

OpenAI is cooking right now with GPT 6 Astra and Image 2.5 in the same week One of the best-selling points that Codex has always had is how strong their built-in image generation is There are literally only three good models with built-in image generation, and that's Codex, MiniMax, and Grok 1. Grok images/videos are straight slop but have no rules and are super cheap 2. MiniMax's H3 video generation is amazing, but images are a bit behind 3. Codex has and always has had the strongest built-in image generation Anthropic doesn't even have built-in image generation, let alone a model worth using nowadays Sketch mode for image 2.5 is also super fun lol it turned this straight-up slop painting of mine into a Demon's Souls boss fight pretty easily lmaoo

这条是引用着另一条写的:x.com/openai/status/2097394956457623964

x.com/ashen_one/status/2097398184897077699采集于 2026.09.14 16:10