gpt-image-1.5
OpenAI$0.165 / imageLatency: 40.9 s
AI IMAGE MODEL COMPARISON · 17 MODELS · SAME 6 PROMPTS
Every text-to-image model on Crazyrouter — GPT Image, Nano Banana, Seedream, Qwen Image, Kling and CogView — answering the same six prompts, side by side. Then run your own prompt against up to four of them with your own API key.
Generated on 2026-09-29 through Crazyrouter, /v1/images/generations, 1024×1024, one shot per model per prompt, no retries beyond the one recorded in gallery.json. A sample, not a benchmark.
An elderly man holding a black umbrella on a rain-soaked city street at night, neon reflections on wet asphalt, 35mm film grain, shallow depth of field, cinematic
夜晚雨后的城市街道上,一位老人撑着黑伞,霓虹灯映在湿漉漉的柏油路上,35mm 胶片颗粒,浅景深,电影感
HTTP 429: Upstream provider capacity or rate limit was reached. Please retry later. (request id: 20260929103424527168534IhY2Lvsq)
A minimalist glass perfume bottle with a gold cap on a pure white background, soft studio lighting, subtle reflection, product photography
纯白背景上一只极简风格的玻璃香水瓶,金色瓶盖,柔和的棚拍灯光,淡淡的倒影,产品摄影
HTTP 429: Upstream provider capacity or rate limit was reached. Please retry later. (request id: 20260929103424570338603im0wuOzi)
A café chalkboard sign that reads "OPEN 7AM – 9PM · Fresh Bagels" in neat white chalk lettering, wooden frame, warm light
一块咖啡馆黑板招牌,用整齐的白色粉笔字写着 "OPEN 7AM – 9PM · Fresh Bagels",木质边框,暖光
HTTP 429: Upstream provider capacity or rate limit was reached. Please retry later. (request id: 20260929103426311377807z7iHUrV9)
一副红纸金字的春联,上面写着“福满人间”四个楷体大字,挂在木门旁,喜庆,高清
HTTP 429: Upstream provider capacity or rate limit was reached. Please retry later. (request id: 20260929103435784947434co4Gcwop)
Three animals — a fox, a raccoon and an owl — playing poker at a round wooden table, each with a distinct expression and prop (monocle, cigar, teacup), warm lamp light, detailed
三只动物——狐狸、浣熊和猫头鹰——围坐在圆木桌旁打扑克,各有不同的表情和道具(单片眼镜、雪茄、茶杯),暖色台灯光,细节丰富
HTTP 429: Upstream provider capacity or rate limit was reached. Please retry later. (request id: 20260929103441272910452usq2adPm)
A small mountain town in a valley at dawn, thin morning mist, Studio Ghibli style illustration, soft colors, hand-painted look
黎明时分山谷中的小镇,薄薄的晨雾,吉卜力工作室风格插画,柔和的色彩,手绘质感
HTTP 429: Upstream provider capacity or rate limit was reached. Please retry later. (request id: 20260929103441292323711ejYMB6oj)
Images were generated by the maintainers through Crazyrouter with the exact request shown in the preview below (/v1/images/generations, 1024×1024, one image per call) and stored as WebP. Prices are per image from /api/pricing at build time; latency is the wall-clock time of that one call. A model that failed that day may succeed tomorrow and vice versa — it is a sample, not a benchmark.
Endpoint, key, prompt, size and models
The key lives only in this page's memory and is sent only to the base URL above. It is never written to browser storage.
Each run sends one real, paid request per model using your own balance. Nothing is retried. Some sizes are rejected by some models with HTTP 400 — that is an upstream limit, not a model failure. Models that answer with a URL instead of image bytes are shown straight from that URL, which usually expires after a while.
One card per model: status, latency, price per image and the generated image
Images returned as base64 are rendered from an in-memory blob URL; URL answers are shown directly. Nothing is uploaded anywhere else.
Same prompt, same size, one shot per model — no cherry-picking, no reruns beyond the single retry recorded in gallery.json for 429/5xx errors. It is a quick visual sense of style, text rendering and prompt adherence across providers, not a benchmark or a quality ranking. All gallery requests were sent with response_format: "b64_json" (or output_format: "png" for gpt-image-*, which reject response_format); models that only return URLs were downloaded server-side. Prices come from Crazyrouter's pricing page at build time and may change.