DALL·E 3
Editorial review编辑测评
2026-10-15Unmatched prompt-following convenience inside ChatGPT, capped by a stock-illustration look and a hard China access wall.听话、省事、能进 API,但出图一股素材库味,且大陆直连与支付双受限。
Bottom line: DALL·E 3 is the most convenient image model ever shipped, and convenience is both its selling point and its ceiling. It lives inside ChatGPT, so the workflow is: describe what you want in plain language, get a decent image. No prompt dialect to learn, no parameters to tune.
Its genuine strength is prompt adherence. Compared with models that treat long descriptions as suggestions, DALL·E 3 actually follows multi-clause instructions: object counts, spatial relations, and on-image text come out roughly where you asked. For product mockups and explainer graphics, that reliability saves more time than a prettier model that needs five rerolls. An official API means it can sit inside real applications, unlike Midjourney.
The weaknesses are aesthetic and geographic. Output defaults to a recognizable stock-illustration look that heavy users learn to spot instantly, and fine artistic control is thin compared with Midjourney or a tuned Stable Diffusion pipeline. Pricing is paid-only via ChatGPT Plus or per-image API calls. Our data flags cn_ok=false, and probes record the site returning 403 from our test region on nearly every check since mid-August — mainland users need both a workaround and a foreign payment method.
Verdict: the right choice when the image is a means to an end and you already pay for ChatGPT; the wrong choice when the image is the product.
结论先行:DALL·E 3 是历史上最省心的画图模型,省心既是卖点也是天花板。它长在 ChatGPT 里,工作流就是说人话、拿图,不用学提示词方言,不用调参数。
真正的强项是指令遵循。面对多分句的复杂描述——物体数量、空间关系、图上文字——它基本按你说的来,而不是当建议听。做产品示意图、讲解配图时,这种确定性比更好看但要重抽五次的模型省时间。官方 API 也让它能进真实应用,这一点胜过 Midjourney,对要把画图嵌进自家产品的团队是决定性差异。
短板在审美和可达性。默认出图一眼 AI 素材库味,重度用户一眼能认;精细艺术控制远不如 Midjourney 或调好的 Stable Diffusion 管线。付费只走 ChatGPT Plus 订阅或按张计费的 API,没有真正意义上的免费档。数据里 cn_ok=false,探活显示官网从大陆测试点几乎天天 403,能连通时延迟约 769ms,大陆用户要同时解决网络和境外支付两道坎。
一句话:图是手段、且已订阅 ChatGPT 时选它;图本身就是产品时,去看别家。
Pros优点
- Best-in-class prompt adherence: multi-clause instructions, object counts, and spatial layouts land as described
- Zero learning curve — plain-language generation inside ChatGPT that non-designers can use on day one
- Official API enables real application integration, unlike API-less rivals such as Midjourney
Cons缺点
- Default aesthetic reads as generic stock illustration; limited fine-grained artistic control
- cn_ok=false with persistent 403s in our probe history; requires ChatGPT Plus or API billing plus a foreign payment method
Best for: 最适合:ChatGPT subscribers who need reliable, instruction-following images for mockups and explainers without learning prompt craft.已在用 ChatGPT、需要稳定听懂指令的配图能力、不愿钻研提示词技巧的用户。
Appears in scenarios出现在这些场景
1 scenarioCompared with与它对比
1 comparisonQuick answers常见问答
Is DALL·E 3 free?DALL·E 3 是免费的吗?
No, DALL·E 3 is paid — see the official site for current pricing.DALL·E 3 付费,具体价格见官网。
Does DALL·E 3 work in China?国内能直接用 DALL·E 3 吗?
DALL·E 3 may require a proxy from mainland China; we recheck its link daily.DALL·E 3 国内可能需要代理;我们每日复测其链接。
Is DALL·E 3's link verified?DALL·E 3 的链接核验过吗?
Yes — we probed DALL·E 3's link 33d ago (HTTP 403, 769 ms) and it's live.是——我们33 天前探测 DALL·E 3 链接(HTTP 403, 769 ms),状态活跃。