跳到主要内容
AI 对话模型
Anthropic
Claude Opus

Claude Opus 5

Claude Opus 5 是高复杂度纯文本任务的旗舰候选,站内限制单条 20,000 字符、最近 40 条历史消息和 4096 最大输出。

定位
新一代旗舰文本
最大输出
4096 tokens
计费
实际积分结算
快速了解
  • 创建前按完整会话最坏用量预授权。
  • 普通与流式都读取实际积分。
  • 纯文本、无附件、最大输出 4096 tokens。

相关查询: Claude Opus 5 怎么用 · Claude Opus 5 价格 · Claude Opus 5 适合什么

示例

Visual prompt study

请整理这段画面创意的生成提示词,保留主体、构图、光线与风格细节。
Design a clever typography logo where the brand name [BRAND NAME] merges together to form the silhouette of a [ANIMAL/OBJECT]. Letters should naturally shape the contours of the figure while maintaining subtle readability. Minimalist vector logo, clean geometry, premium branding aesthetic.

Visual prompt study

请整理这段画面创意的生成提示词,保留主体、构图、光线与风格细节。
A premium studio product photograph of a tall, slim red aluminum sparkling water can standing upright and centered against a seamless vivid red background, the can covered in fine condensation droplets with crisp white typography reading “GOOD IDEA” and smaller flavor text beneath, positioned on a matte red surface with subtle reflections, surrounded symmetrically by fresh ripe strawberries with glossy seeds and deep red flesh, some resting directly on the surface and others placed on small red rectangular blocks to the left and right, delicate white elderflower blossoms scattered between the fruit for contrast, the lighting soft yet directional from the upper left creating gentle highlights on the can's metallic texture and soft shadows beneath the fruit, the color palette dominated by saturated reds with clean white accents and natural green stems, the composition minimal and perfectly balanced with generous negative space above the product, shallow depth of field, ultra-sharp focus on the can, high-end commercial aesthetic, photorealistic, studio-grade, cinematic realism.

Visual prompt study

请整理这段画面创意的生成提示词,保留主体、构图、光线与风格细节。
一位年轻成年东亚女性,纤细清秀,坐姿近景到半身肖像。她上半身微微前倾,身体略向右侧,头部轻微倾斜,双眼温柔地看向镜头,神情安静、柔和、克制,带一点自然亲近感。她有白皙偏暖的肤色,真实细腻的皮肤质感,保留毛孔、轻微绒毛、自然阴影和柔和高光,不要塑料磨皮。五官精致,杏仁眼,柔和眉形,小巧鼻梁,自然裸色嘴唇,淡妆。 发型为深黑或深棕黑色头发,松散盘成高位凌乱丸子头,带自然碎发和轻薄空气刘海,耳侧和颈侧有几缕散落发丝,呈现轻盈随性的居家感。 服装是一件白色细肩带吊带上衣,轻薄柔软,带细腻蕾丝或镂空纹理细节,领口低调自然,不过分夸张,整体清爽、纯净、简约。前景可见一只弯曲抬起的膝盖,形成柔和的近景遮挡和亲密构图。 背景是极简的白色或浅灰白室内空间,无明显杂物,干净、通透、安静。光线为高键柔光,像窗边自然散射光或柔和棚拍漫射光,从左前方轻柔照亮面部和肩颈,阴影极浅,整体低对比、低饱和,呈现奶油白、浅灰、裸粉肤色的高级日系空气感。 摄影风格:85mm 人像镜头,浅景深,焦点清晰落在眼睛和面部,背景轻柔虚化,真实摄影质感,唯美但不假,干净克制,high-key soft portrait,editorial beauty portrait,photorealistic。 避免: 卡通感、CG感、过度磨皮、浓妆、夸张表情、杂乱背景、错误手指、畸形手臂、过强锐化、过重滤镜、文字、水印、Logo。

Visual prompt study

请整理这段画面创意的生成提示词,保留主体、构图、光线与风格细节。
Please create a separate "Rubber Stamp Travel Field Notes Poster" for each photo I upload, outputting each photo individually without collage or multi-image combinations. Overall, use a 4:3 landscape composition, dividing the frame into left and right regions, but without drawing an obvious dividing line. The left side takes up about 58% of the frame, faithfully preserving the original photo. Accurately maintain the main subject identity, terrain, architecture, plants, people, spatial relationships, natural lighting and shadows, authentic textures, and the original color atmosphere; apply only restrained art publication-level photo color grading, and add extremely subtle, fine-grained film noise. For layout adaptation, natural cropping is allowed, but do not stretch, distort, shift, replace, or redraw the main subject. The right side takes up about 42% of the frame, using a warm off-white aged paper as the background. The paper features subtle fibers, natural grain, light usage marks, and a matte texture, while preserving large areas of unprinted paper whitespace, making the blank space an essential part of the layout. Analyze the original photo and extract the most location-distinctive subject outlines, architectural structures, terrain contours, plant forms, roads, shorelines, or other key visual relationships, compressing them into a small multi-color rubber stamp image. Do not replicate every single element from the photo item by item. Retain only the minimal information necessary to instantly recognize the original location, subject, and scene relationships. Remove crowds, vehicles, dense windows, repetitive buildings, fragmented vegetation, decorative elements, and irrelevant backgrounds. The stamp is positioned in the lower-middle of the right-side paper area, occupying only about 30%–38% of the right region's height, with ample whitespace preserved around it. The stamp must not be enlarged into a standard illustration, full landscape painting, or brand logo. Determine the stamp's organization based on the original photo's composition: * Iconic architecture: Retain the most distinctive outer contours, roofs, domes, arches, towers, or main structures. * Mountain settlements: Compress buildings into a few terraced color blocks aligned along the terrain. * Coastal scenery: Retain mountain contours, settlement layers, shorelines, and sparse intermittent water ripples. * City panoramas: Retain the main skyline, one iconic building, and one or two layers of distant mountains. * Natural landscapes: Retain primary mountain forms, trees, shorelines, or road orientations. * Foreground occlusions: If narratively important in the original photo, retain as foreground stamp outlines. Extract 2–4 spot inks from the original photo. Prioritize desaturated colors like carbon black, deep green, brick red, ochre yellow, slate blue, or taupe brown, but do not force a fixed palette. Preserve the most distinctive color character from the original photo, allowing only a small area of color for visual emphasis. Render each color as a separately hand-stamped effect: Authentic rubber stamp carving texture, hand-engraved marks, uneven line widths, contour notches, fractured edges, dry ink shortages, paper show-through, granular ink, uneven pressure, partial ghosting, and about 1–2 mm of subtle misregistration. Allow natural misalignment between color layers; edges must not be digitally smoothed. The print should resemble a real carved stamp pressed onto aged paper, not a filtered photo, smooth vector illustration, or line-art logo. Generate text based on the photo's location, theme, and visual imagery: Location English name No. Number Three short English keywords Gregorian calendar year Place the text below or adjacent to the stamp in the whitespace, using a small, restrained, slightly mechanically imperfect typewriter font. The typography should evoke a traveler's field record, not an ad headline. Ensure all text is spelled accurately, without adding irrelevant slogans, brands, or decorative copy. The overall vibe is like field notes kept by an architect, travel writer, or natural observer: quiet, restrained, tactilely real, regionally specific, with handmade imperfections and a collectible feel. The photo handles the on-site record; the stamp captures the most recognizable fragments of memory. The date should display the date I enter: [Date] Avoid: Obvious central dividing lines, circular seals, Chinese red stamps, postage stamp perforations, wax seals, sticker collages, tourist souvenir templates, smooth vector logos, generic city icons, full replication of all architecture, dense detailing, childlike craftiness, cartoon style, 3D rendering, plastic textures, glossy digital gradients, oversaturation, excessive text, decorative clutter, and redrawing or altering the left-side original photo.

Visual prompt study

请整理这段画面创意的生成提示词,保留主体、构图、光线与风格细节。
Create a clean premium logo design for [Brand Name] based on the letter [Core Letter]. Turn the letter into an optical illusion brand mark using folded geometry, spatial depth, precise edges and strong negative space. The logo should express [Symbol Meaning] while remaining readable, minimal and memorable. Center one large logo mark on a plain warm ivory or soft light gray background. Add only the brand name “[Brand Name]” underneath, with an optional short subtitle “[Subtitle]”. Minimal black-white-gray palette, premium modern identity design. No poster layout, no construction grid, no annotations, no extra icons, no decorative small text, no mockup clutter, no messy 3D, no distorted typography.

Visual prompt study

请整理这段画面创意的生成提示词,保留主体、构图、光线与风格细节。
Creative composite photograph, 9:16 vertical aspect ratio, clean minimalist white background, studio lighting, high detail, sharp focus. A real-life [EVERYDAY OBJECT] is physically placed and perfectly integrated as a key part of a hand-drawn black ink cartoon illustration. The drawing is simple, expressive line art with minimal shading, like a quick sketch on paper. The [OBJECT] represents [CLEVER PART OF THE SCENE - e.g. hair, weapon, body part, clothing, etc.], cleverly completing the visual pun or concept. The cartoon character interacts dynamically with the object in a humorous, meaningful, or surprising way. Photorealistic object + bold black line drawing style, perfect seamless blend between real object and illustration, clean negative space, artistic and witty concept, high quality product photography mixed with doodle art. • How to edit super easy Just replace the placeholders: [EVERYDAY OBJECT] → what real thing you'll use (tomato, onion, chips, etc.) [CLEVER PART OF THE SCENE] → what the object becomes in the drawing Then customize the character and concept

一句话了解

这个模型怎么样?

模型创建前会依据完整消息的保守输入上界与最大输出预授权;成功后按实际积分结算。它不接受附件,也不把其他 Claude 入口的视觉或工具能力视为本站能力。

模型定位
面向复杂推理、严谨分析和高价值专业输出的 Claude 旗舰。
适合谁用
多约束决策、关键文档审阅、研究综合与需要降低返工风险的任务。
本站使用说明
通过明确的 4096 输出上界实现可靠预授权,并保留实际积分对账证据。

关键参数一览

根据这些信息,判断模型是否适合你的任务。

厂商
Anthropic
家族
Claude Opus
输入
纯文本
最大输出
4096 tokens
历史消息
最近 40 条已提交消息

它还叫什么

同一个模型在公开资料中可能有不同名称,这里整理相关称呼,方便查找和比较。

Claude Opus 5 Chat
Opus 5

用户常见问题

从任务、输入条件和结果要求出发,整理选择模型时需要确认的问题。

Claude Opus 5 怎么用
Claude Opus 5 价格
Claude Opus 5 适合什么
Claude Opus 5 和 Opus 4.8 区别

选型指南

根据输入素材、任务要求和积分成本,逐项确认模型是否适合。

1
结果影响重大

当错误会带来高返工或决策成本时,用代表性样本比较 Opus 5。

2
常规知识工作

先使用费率更低的 Sonnet 5,只有明显收益时再升级。

3
需要长交付

拆为提纲、章节和最终审阅,避免单次触及输出上限。

模型对比

把常见候选放在同一任务下比较,帮助你确认输入方式、控制项和成本差异。

Opus 5 与 Sonnet 5 怎么选
Opus 5 是否支持图片

适用情况

依据公开资料与当前站内设置,说明模型适用的任务及需要注意的限制。

旗舰选择需要任务级评价

关注正确性、遗漏率、返工轮次与最终可用性,而非只比较文风。

结算证据是上线能力的一部分

高费率模型需要创建前覆盖和完成后实际用量两端都可审计。

能力详解

约束推理

同时考虑目标、限制、反例和风险。

关键审阅

检查材料中的冲突、缺口和无证据结论。

专业写作

按受众和决策场景组织高密度文本。

适合做什么

战略方案

比较备选路线并列出条件、代价和验证步骤。

高风险文档初审

发现逻辑或证据问题,最终仍由专业人员确认。

研究报告

把材料整理为可追溯的论点与行动建议。

提示词技巧

明确证据范围

列出只能使用的材料与必须标记的未知项。

要求反方检查

让输出主动寻找最强反例和失败条件。

先短样本再扩展

确认质量和实际积分后再处理更长材料。

为什么选它

旗舰复杂推理定位。
预授权和结算链路完整。
运行边界不依赖猜测。

使用前要知道

不接受图片或文件附件。
单次输出最多 4096 tokens。
高风险结论不能替代专业审核。

可调参数

根据这些信息,判断模型是否适合你的任务。

当前页面未列出额外可调设置,请以工作台为准。

积分消耗

按 tokens 计费

输入 400 积分 / 百万 tokens,输出 2000 积分 / 百万 tokens;按实际返回积分结算。

使用建议

准备多次生成时,建议先用少量素材测试,核对输入设置和结果是否适合任务,再决定是否继续。每次生成均按工作台显示的计费规则扣点。

常见问题

相关模型

按你的任务比较这些模型的输入要求、功能和当前状态,再查看对应详情。

Claude Opus 4.8

Claude Opus 4.8 面向高复杂度文本分析与关键写作;本站开放纯文本对话,单条最多 20,000 字符,并显式限制最大输出为 4096 tokens。

AI 对话模型
按 tokens 计费

Claude Sonnet 5

Claude Sonnet 5 是本批 Claude 中更均衡的文本候选,适合研究、代码解释与结构化写作;站内最大输出为 4096 tokens。

AI 对话模型
按 tokens 计费

Claude Fable 5

Claude Fable 5 面向高要求的叙事构建与文本创作协作;本站只开放纯文本,最大输出 4096 tokens,不发送工具或附件字段。

AI 对话模型
按 tokens 计费

信息来源

页面内容依据公开模型资料与 Kyeo AI 的模型信息整理。能否在线使用,请以当前模型状态为准。

来源说明

Claude Opus 5 已按纯文本 Messages 合同接入;显式发送 4096 最大输出,并让普通与流式响应沿用实际积分三分支结算。

内容复核日期: 2026-08-11
已锁定的 Claude Opus 5 接入合同
平台

记录纯文本 Messages、4096 最大输出、费率与实际积分结算。