AI 模型库

Seedance 2.5 text-to-video with 4-30s output, 480p/720p/1080p, synchronized audio, and optional web search.

Text-to-video generation with optional web search. 4-15s duration, 480p/720p/1080p, native audio sync.

Lightweight text-to-video generation. 4-15s duration, 480p/720p, native audio sync.

OpenAI 高级图像生成模型，优化真实色彩精度、结构化任务与分析型视觉输出，基于 DALL·E 3 技术。

Gemini Omni 1.1 Flash text-to-video with 3-10s output, 360p/720p/1080p/4K, and natively synchronized audio.

Tongyi Wanxiang 3.0 video generation with text-to-video, image-to-video, and all-in-one reference-to-video. 480p/720p/1080p output, 2-30s or smart duration, per-second billing.

Tongyi Wanxiang 3.0 Prime is the speed-oriented tier for significantly faster end-to-end video generation. Use three EvoLink model IDs for text, image, or all-in-one reference workflows up to 30 seconds.

通过 EvoLink 统一 API 调用 xAI 图像生成与编辑，支持 1-3 张参考图、1K/2K 输出、13 种比例，单次最多生成 10 张。

MiniMax H3 API，提供文生视频、图生视频（首尾帧）和参考素材生视频三条路由。输出 2K / 768p、时长 4-15 秒，按输出秒计费。

智谱面向编程 Agent 与长周期工程任务的旗舰推理模型。100 万 Token 上下文、思考常开且分三档强度、提示缓存，Chat Completions 与 Anthropic Messages 双端点。

5.3 这一代的原生多模态走量档。文本、图片、视频、文件输入，100 万 Token 上下文，价格约为 GLM-5.3 的九分之一。

Native multimodal model: text and image input with 1M context window. Same pricing as V4 Flash.

xAI 最新推理与工具调用模型：500,000 Token 上下文、Chat Completions 与 Responses 双端点、缓存输入，以及五类按次计费的服务端工具。

Advanced reasoning model with thinking mode and 1M context window. Strongest DeepSeek tier.

Fast general-purpose model with 1M context window. Optional thinking mode.

Anthropic 最新的 Opus 旗舰模型，面向仓库级编程、长周期智能体与高风险审查。100 万上下文，12.8 万最大输出。

xAI Grok Imagine Video 1.5 Preview: text-to-video, image-to-video, and reference-to-video in one route — the number of input images (0, 1, or 2-7) selects the mode. 1-15s duration, 480p/720p/1080p.

Advanced Seedream 5.0 Pro image generation and editing with 1K/1.5K/2K output tiers, up to 10 reference images, and unified asynchronous task responses.

Tongyi Wanxiang image generation & editing 3.0 (DashScope). Text-to-image and image-to-image editing with up to 3 reference images, 1-6 outputs, flexible output sizes (auto, 1K/2K, aspect ratio, or custom pixels up to ~4.19MP), and smart prompt rewriting. Output billed per generated image with 1K and 2K priced the same; reference images billed separately.

Lite variant of Nano Banana 2 (Gemini 3.1 Flash-Lite Image): ~4s generation (2.7× faster than Nano Banana 2), native 1K output, and 14 aspect ratios. The fastest, lowest-cost tier for high-volume image generation and editing.

xAI 推理与工具调用模型，提供 500,000 Token 上下文、Chat Completions、Responses、缓存输入与按次计费的服务端工具。

Moonshot 新一代推理模型，提供 1,048,576 Token 上下文，并支持 Chat Completions 与 Messages 双协议。

Google 面向编程与智能体的 Flash 档主力模型；100 万 Token 上下文，代码生成与终端执行强于 3.6 Flash，支持提示词缓存与文本/图像/视频/音频输入。

Google's fast multimodal Flash model with a 1M context window, built-in reasoning, prompt caching, and unified text/image/video/audio input pricing.

Google's lowest-cost Flash Lite model with a 1M context window and multimodal input, built for high-volume, latency-sensitive, and batch workloads.

Gemini Omni Flash 视频生成与编辑模型，支持文生视频、图生视频、参考图生视频和视频编辑，统一走视频 API 端点。

新一代 Nano Banana 模型，带来更高质量的图像生成、更强细节和出色的提示词遵循。兼顾速度与成本效率，是面向创意专业人士的重大升级。

OpenAI 推理模型家族，提供 Sol、Terra、Luna 三个档位，支持 1.05M 上下文、128K 最大输出与提示词缓存。

Prompt-based AI audio generation for voice, dialogue, sound effects, music, and ambience, with optional reference audio guidance. Per-second billing.

Anthropic 面向编程与 Agent 任务最强的 Sonnet；1M 上下文，128K 最大输出，默认开启 adaptive thinking。

Midjourney V8.1 generates 4 images per request with HD/2K output options, reference-image workflows, and Draft/Fast speed tiers. Supports text-to-image and image-to-image through EvoLink.

高性价比 Nano Banana 模型，输入图像免费。适合大批量生成，质量/价格比出色。

Premium text-to-image model with selectable 1K and 2K output. Built for crisp typography, cinematic lighting, and high-fidelity poster-grade compositions.

Kling 3.0 Turbo text-to-video — faster generation at 720P/1080P. Supports 3-15 second videos with per-second billing.

Z.ai's GLM-5.2 flagship text model for coding agents and agentic workflows, with a ~1M context window, deep thinking, tool calling, and prompt caching. Served over an OpenAI-compatible /v1/chat/completions endpoint.

Text-to-video generation. 3-15s duration, 720p/1080p, 9 aspect ratios, per-second billing.

Anthropic's most powerful and most intelligent Claude model — a new tier above Opus, with state-of-the-art reasoning, coding, and agentic capabilities. 1M context window.

Tongyi Wanxiang 2.7 video generation model with text-to-video, image-to-video, reference video, and video editing variants.

Text-to-video generation. 3-15s duration, 720p/1080p, per-second billing.

MiniMax's flagship multimodal text model with ~1M context, deep thinking, image/video/PDF input, and prompt caching. Available on both OpenAI-compatible (/v1/chat/completions) and Anthropic Messages (/v1/messages) endpoints.

Anthropic's most powerful Claude model with exceptional reasoning, coding, and agentic capabilities. 1M context window.

Google's next-gen Flash model with multimodal input (text/image/video/audio) at unified price and built-in reasoning output

阿里通义千问旗舰 Max 模型，提供百万 Token 上下文、可控思考与提示词缓存，支持三种 API 协议。

OpenAI 最新旗舰模型，具备先进推理能力；支持 1M 上下文、128K 最大输出与 Tool Search 功能。

Midjourney V7 generates 4 stunning images per request with native MJ prompt syntax. Supports text-to-image and image-to-image with three speed tiers (Draft/Fast/Turbo).

Multimodal content safety classifier for text and images. Detects 13 categories of harmful content (harassment, hate, sexual, violence, self-harm, illicit). OpenAI-compatible /v1/moderations endpoint.

AI-powered video super resolution. Enhance video quality with 1x, 2x, or 4x upscaling. Supports MP4 up to 50MB.

新一代 AI 图像生成，支持 2K/3K 画质、联网搜索增强、自定义像素尺寸。支持批量生成（1–15 张）。

支持文生视频与图生视频，音频可选。时长 4–12 秒，480p/720p/1080p 画质。

Google's most cost-efficient model for high-volume agentic tasks, translation, classification, and data processing

Transfer human motion from a reference video onto a character in a reference image. Supports std/pro quality with per-second billing.

Kling O3 (V3 Omni) next-generation video model with text-to-video, image-to-video, reference-to-video, and video editing. Supports 3-15 second videos with per-second billing.

Kling 3.0 video model with text-to-video and image-to-video. Supports 3-15 second videos with per-second billing.

通义万相 2.6 视频模型，提供文生视频、图生视频与参考视频版本。

Google DeepMind 新一代视频模型，含 Fast/Pro 版本；支持 8 秒视频生成，画质更强。

OpenAI 最新 10–15 秒视频生成模型，含音频。支持去水印（价格 1.65×）。

超高速图像生成模型，质量与速度比出色。适合高并发大批量生成，提示词遵循优秀。

先进 AI 生图，支持 2K/4K 画质；支持批量生成（1–15 张）、参考图与灵活尺寸。

xAI Grok Imagine 视频生成 API，支持文生视频和图生视频。6-30 秒任意时长，含 fun/normal/spicy 风格模式。

AI 音乐生成，支持人声、歌词与伴奏。多版本模型，通过文本提示生成专业级歌曲。

Anthropic's most powerful Claude model with exceptional reasoning, coding, and agentic capabilities. 1M context window.

Google's advanced model optimized for custom tool calling and function execution with full reasoning capabilities

MiniMax's latest text model deployed on Alibaba Cloud, excelling at coding, office tasks, and text summarization with fast output speed. 204K context window with built-in reasoning capabilities.

OpenAI 最新旗舰模型，面向编程与 Agent 任务；支持 1.05M 上下文、128K 最大输出与高级推理能力。

Google's latest iteration of Gemini 3 Pro with advanced multimodal capabilities and extended context support

编程与 Agent 任务的最佳均衡之选，兼顾速度、智能与成本；200K 上下文，128K 最大输出，支持延伸思考。

BytePlus's latest LLM series with 256K context, tiered pricing by prompt length (32K/128K/256K), and cache billing. Available in Pro, Lite, Mini, and Code variants.

High-performance general-purpose chat model (DeepSeek-V3) with 128K context window and competitive pricing for everyday AI tasks

Advanced reasoning model (DeepSeek-R1) with chain-of-thought capabilities, 128K context window, optimized for complex problem-solving tasks

Anthropic's most powerful Claude model with exceptional reasoning, coding, and agentic capabilities. 1M context window.

Google's most cost-efficient model for high-volume tasks like translation, classification, and data processing

通义万相图像模型（Wan 2.5 Image），提供文生图与图生图版本。

通义万相视频模型（Wan 2.5 Video），包含图生视频与文生视频版本。

OpenAI 高级图像生成模型，优化真实色彩精度、结构化任务与分析型视觉输出，基于 DALL·E 3 技术。

Google 速度最快的前沿模型，3× 速度提升，多模态含音频输入，价格更具性价比。

MiniMax Hailuo 2.3 API，含 Fast/Standard 版本，支持 T2V/I2V，输出 768p/1080p。

MiniMax Hailuo 02 全功能版，支持 T2V/I2V/FLF 模式，分辨率 512p/768p/1080p。

OpenAI 旗舰模型，面向编程与 Agent 任务；支持 400K 上下文、128K 最大输出与高级推理能力。

OpenAI 最新旗舰模型，具备高级推理、Prompt 缓存与 400K 上下文，适合复杂任务。

Anthropic 最强 Claude 模型，具备卓越的推理、编程与 Agent 能力；200K 上下文。

Kling O1 视频模型，含图生视频、视频编辑与快速编辑版本。支持 3–20 秒视频，可用参考图进行风格引导生成。

Google 新一代语言模型，具备高级多模态能力与更长上下文支持。

快速且高性价比的编程助手，200K 上下文，支持 Prompt 缓存以优化性能。

面向 Agent 构建与编程的最强模型，支持 200K 上下文、延伸思考与高级推理。

Google 高速高效语言模型，支持 Prompt 缓存以优化成本。

先进视频模型，支持文生视频与图生视频。时长 2–12 秒，720p/1080p 画质可选。

Google 最强语言模型，支持更长上下文与高级推理能力。

AI 数字人视频生成，支持音频驱动口型同步。可将静态图像变为自然表情与动作的会说话虚拟人。

通义千问图像编辑模型，具备智能理解与多图协同编辑能力。

Gemini 2.5 Flash Image Preview 是先进 AI 模型，擅长自然语言驱动的图像生成与编辑。

叙事型 AI 生图，支持 4K 质量：多参考融合与实时编辑，可生成 9+ 张一致视觉。

OpenAI named Astra as its next major model but has not linked it to GPT-6 or published an API. Track verified model IDs, pricing, limits, and EvoLink route readiness.

Claude Fable 5.1 尚未官宣。可在这里追踪正式产品名称、API 接入、价格、安全限制与 EvoLink 路由验证状态。

追踪第三方报道中的发布进展、已验证的 EvoLink API 可用性、模型 ID、价格与编程 Agent 评测准备。

Gemini 3.5 Pro is testing with partners, but no public API route, model ID, pricing, or input specification is available yet.