如何调用模型
使用AISA_API_KEY 作为 Bearer token。对于 OpenAI 兼容 SDK，将 base URL 设置为 https://api.aisa.one/v1。
import os
from openai import OpenAI
client = OpenAI(
api_key=os.environ["AISA_API_KEY"],
base_url="https://api.aisa.one/v1"
)
response = client.chat.completions.create(
model="qwen3.7-max",
messages=[{"role": "user", "content": "Compare these model options for a coding agent."}]
)
print(response.choices[0].message.content)
Endpoint 类型
| Endpoint | 当前模型数 |
|---|---|
POST /v1/chat/completions | 101 |
POST /v1/messages | 89 |
POST /v1/responses | 80 |
POST /v1/video/generations | 9 |
POST /v1/images/generations | 6 |
POST /v1/images/edits | 5 |
POST /v1/embeddings | 4 |
POST /v1beta/models/{model}:generateContent | 1 |
POST /v1/classify | 1 |
POST /v1/rerank | 1 |
POST /v1/systemone | 1 |
chat/completions 和 Anthropic 兼容的 messages，因此同一个模型 ID 在两种 SDK 下都能直接用。POST /v1/responses 现已覆盖大多数文本模型 —— 上表合并统计了 2026 年 8 月 5 日元数据导出中的路由和 2026 年 9 月 30 日线上网关定价数据报告的路由。线上数据可能漏报模型实际支持的路由，因此新增模型的 Endpoint(s) 列可能缺少 messages 或 responses。在模型不支持的路由上调用会返回 400 model_route_not_supported，并附带该模型支持的路由列表，因此切换协议前请先探测。Gemini 模型额外提供 generateContent。图像模型使用 images/generations 和 images/edits。视频模型是异步任务制：先提交到 POST /v1/video/generations，再轮询 /v1/video/generations/{task_id} 获取结果。Embedding 模型使用 /v1/embeddings，Jina reranker 使用 /v1/rerank。调用时请始终使用下表中的精确模型字符串。
能力标签说明
| 能力 | 在 AIsa Model Gateway 中的含义 |
|---|---|
| 文本 | 自然语言生成、摘要、分析、翻译和长上下文推理。 |
| 编程 | 代码推理、代码补全、软件 Agent 规划和工具调用流程。 |
| 视觉 | 图像/文档理解、视觉编码，以及基于视觉输入的空间推理。 |
| 音频 | 上游模型支持时的音频理解或语音到语音交互。 |
| 图像 | 图像生成、图像编辑、图像一致性，或在图像中渲染文字。 |
| 视频 | 视频理解、时序推理、长视频处理、图生视频或全模态/视频流程。 |
Provider 概览
| Provider | 模型数 | 类型 | 示例模型 ID |
|---|---|---|---|
| OpenAI | 29 | Embeddings, Image, Text | gpt-4.1-mini, gpt-4o, gpt-4o-mini, gpt-5-nano … |
| Anthropic | 13 | Text | claude-fable-5, claude-fable-5-1, claude-haiku-4-5-20251001, claude-opus-4-5-20251101 … |
| Google Gemini | 1 | Text | gemini-3.5-flash |
| xAI | 7 | Text | grok-4.20-0309-non-reasoning, grok-4.20-0309-reasoning, grok-4.3, grok-4.5 … |
| DeepSeek | 10 | Text | deepseek-r1, deepseek-v3, deepseek-v3.1, deepseek-v3.2 … |
| Alibaba | 24 | Image, Text, Video | qwen-flash, qwen-mt-flash, qwen-mt-lite, qwen-plus-2025-12-01 … |
| Moonshot | 5 | Text | kimi-k2-thinking, kimi-k2.5, kimi-k2.6, kimi-k2.7-code … |
| MiniMax | 4 | Text | MiniMax-M2.5, MiniMax-M2.7, MiniMax-M2.7-highspeed, MiniMax-M3 |
| Zhipu GLM | 6 | Text | glm-5, glm-5.1, glm-5.2, glm-5.3 … |
| ByteDance | 10 | Image, Text, Video | dreamina-seedance-2-0-260128, dreamina-seedance-2-0-fast-260128, dreamina-seedance-2-5, seed-1-6-250915 … |
| Tencent | 2 | Text | hy3, hy4-preview |
| Xiaomi | 3 | Text | mimo-v2.6-flash, mimo-v2.6-pro, mimo-v2.6-pro-ultraspeed |
| StepFun | 3 | Text | step-3.5-flash, step-3.5-flash-2603, step-3.7-flash |
| HappyHorse | 3 | Video | happyhorse-1.1-i2v, happyhorse-1.1-r2v, happyhorse-1.1-t2v |
| Jina | 3 | Embeddings | jina-embeddings-v3, jina-embeddings-v5-text-small, jina-reranker-v3 |
| TypeSafe | 1 | Text | jev-latest |
完整模型详情
上下文窗口和能力标签来自上一次 Model Gateway 元数据导出。此后新增的模型，上下文列显示—，能力列只标注基础类型 —— 这些模型的实时上下文限制和能力标签请查看 aisa.one/models。
OpenAI
| Model ID | 上下文 | 能力 | Endpoint(s) | 计费方式 |
|---|---|---|---|---|
gpt-4.1-mini | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
gpt-4o | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
gpt-4o-mini | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
gpt-5-nano | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
gpt-5-pro | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出 |
gpt-5-search-api | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
gpt-5.1 | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
gpt-5.2 | 400,000 | Coding, Text, Vision; reasoning, long context, translation, creative writing, spatial vision, document vision, code reasoning, code completion, agentic coding | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
gpt-5.2-chat-latest | 400,000 | Coding, Text, Vision; reasoning, long context, translation, creative writing, spatial vision, document vision, visual coding, code reasoning, code completion, agentic coding | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
gpt-5.2-pro | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出 |
gpt-5.3-chat-latest | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
gpt-5.3-codex | 1,000,000 | Coding, Text; reasoning, long context, code reasoning, code completion, agentic coding | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
gpt-5.4 | 1,050,000 | Coding, Text, Vision; reasoning, long context, translation, creative writing, spatial vision, document vision, visual coding, code reasoning, code completion, agentic coding | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
gpt-5.4-mini | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
gpt-5.4-nano | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
gpt-5.4-pro | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出 |
gpt-5.5 | 400,000 | Coding, Text, Vision; code reasoning, long context, reasoning, vision | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
gpt-5.5-pro | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出 |
gpt-5.6-luna | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取、缓存写入 |
gpt-5.6-sol | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取、缓存写入 |
gpt-5.6-terra | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取、缓存写入 |
gpt-6-astra | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取、缓存写入 |
gpt-6-luna | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取、缓存写入 |
gpt-6-sol | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取、缓存写入 |
gpt-image-2 | 不适用 | Image, Vision; image editing, image generation, text in images, vision | /v1/images/generations, /v1/images/edits | 按 1M Token：输入、输出、缓存写入；按次请求；按张图片，尺寸分档 |
gpt-image-2.5-flare | — | Image | /v1/images/generations, /v1/images/edits | 按 1M Token：输入、输出、缓存读取 |
gpt-image-2.5-sunburst | — | Image | /v1/images/generations, /v1/images/edits | 按 1M Token：输入、输出、缓存读取 |
text-embedding-3-large | — | Embeddings | /v1/embeddings | 按 1M Token：输入 |
text-embedding-3-small | — | Embeddings | /v1/embeddings | 按 1M Token：输入 |
Anthropic
| Model ID | 上下文 | 能力 | Endpoint(s) | 计费方式 |
|---|---|---|---|---|
claude-fable-5 | — | Text | /v1/chat/completions, /v1/messages | 按 1M Token：输入、输出、缓存读取、缓存写入 |
claude-fable-5-1 | — | Text | /v1/messages | 按 1M Token：输入、输出、缓存写入 |
claude-haiku-4-5-20251001 | 200,000 | Coding, Text, Vision; reasoning, long context, spatial vision, document vision, visual coding, code reasoning, agentic coding | /v1/chat/completions, /v1/messages | 按 1M Token：输入、输出、缓存读取、缓存写入 |
claude-opus-4-5-20251101 | 1,000,000 | Coding, Text, Vision; reasoning, long context, creative writing, document vision, visual coding, code reasoning, code completion, agentic coding | /v1/chat/completions, /v1/messages | 按 1M Token：输入、输出、缓存读取、缓存写入 |
claude-opus-4-6 | 1,000,000 | Coding, Text, Vision; reasoning, long context, creative writing, document vision, visual coding, code reasoning, code completion, agentic coding | /v1/chat/completions, /v1/messages | 按 1M Token：输入、输出、缓存读取、缓存写入 |
claude-opus-4-7 | 1,000,000 | Coding, Text, Vision; reasoning, long context, creative writing, spatial vision, document vision, visual coding, code reasoning, code completion, agentic coding | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取、缓存写入 |
claude-opus-4-8 | 1,000,000 | Coding, Text, Vision; reasoning, long context, creative writing, spatial vision, document vision, visual coding, code reasoning, code completion, agentic coding | /v1/chat/completions, /v1/messages | 按 1M Token：输入、输出、缓存读取、缓存写入 |
claude-opus-5 | — | Text | /v1/chat/completions, /v1/messages | 按 1M Token：输入、输出、缓存读取、缓存写入 |
claude-opus-5-5 | — | Text | /v1/messages | 按 1M Token：输入、输出、缓存读取、缓存写入 |
claude-sonnet-4-5-20250929 | 200,000 | Coding, Text, Vision; reasoning, long context, creative writing, spatial vision, document vision, visual coding, code reasoning, code completion, agentic coding | /v1/chat/completions, /v1/messages | 按 1M Token：输入、输出、缓存读取、缓存写入 |
claude-sonnet-4-6 | — | Text | /v1/chat/completions, /v1/messages | 按 1M Token：输入、输出、缓存读取、缓存写入 |
claude-sonnet-5 | — | Text | /v1/chat/completions, /v1/messages | 按 1M Token：输入、输出、缓存读取、缓存写入 |
claude-sonnet-5-5 | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取、缓存写入 |
Google Gemini
| Model ID | 上下文 | 能力 | Endpoint(s) | 计费方式 |
|---|---|---|---|---|
gemini-3.5-flash | 不适用 | Text; long context, creative writing | /v1/chat/completions, /v1beta/models/{model}:generateContent | 按 1M Token：输入、输出、缓存读取 |
xAI
| Model ID | 上下文 | 能力 | Endpoint(s) | 计费方式 |
|---|---|---|---|---|
grok-4.20-0309-non-reasoning | 1,000,000 | Text, Vision; long context, creative writing, spatial vision, document vision | /v1/chat/completions | 按 1M Token：输入、输出 |
grok-4.20-0309-reasoning | 1,000,000 | Text, Vision; reasoning, long context, creative writing, spatial vision, document vision | /v1/chat/completions | 按 1M Token：输入、输出 |
grok-4.3 | 1,000,000 | Text, Vision; reasoning, long context, creative writing, spatial vision, document vision | /v1/chat/completions | 按 1M Token：输入、输出 |
grok-4.5 | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存写入 |
grok-4.6 | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
grok-4.7 | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
grok-build-0.1 | 256,000 | Coding, Text, Vision; reasoning, long context, visual coding, code reasoning, code completion, agentic coding | /v1/chat/completions | 按 1M Token：输入、输出 |
DeepSeek
| Model ID | 上下文 | 能力 | Endpoint(s) | 计费方式 |
|---|---|---|---|---|
deepseek-r1 | 262,144 | Coding, Text; code reasoning, long context, reasoning | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
deepseek-v3 | 262,144 | Coding, Text; code reasoning, long context, reasoning | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
deepseek-v3.1 | 262,144 | Coding, Text; code reasoning, long context, reasoning | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
deepseek-v3.2 | 128,000 | Coding, Text; reasoning, long context, creative writing, code reasoning, code completion, agentic coding | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
deepseek-v4-flash | 262,144 | Coding, Text; code reasoning, long context, reasoning | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
deepseek-v4-flash-0731 | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存写入 |
deepseek-v4-flash-vision-exp | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存写入 |
deepseek-v4-pro | 262,144 | Coding, Text; code reasoning, long context, reasoning | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
deepseek-v4-pro-0813 | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存写入 |
deepseek-v4.1-flash | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存写入 |
Alibaba
| Model ID | 上下文 | 能力 | Endpoint(s) | 计费方式 |
|---|---|---|---|---|
qwen-flash | 1,000,000 | Audio, Coding, Text, Video, Vision; reasoning, long context, translation, creative writing, spatial vision, document vision, visual coding, speech-to-speech, omni/video understanding, long video, temporal video, code reasoning, code completion, agentic coding | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取、缓存写入 |
qwen-mt-flash | 1,000,000 | Text; long context, translation | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
qwen-mt-lite | 1,000,000 | Text; translation | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
qwen-plus-2025-12-01 | 1,000,000 | Audio, Coding, Text, Video, Vision; reasoning, long context, translation, creative writing, spatial vision, document vision, visual coding, speech-to-speech, omni/video understanding, long video, temporal video, code reasoning, code completion, agentic coding | /v1/chat/completions | 按 1M Token：输入、输出、缓存读取 |
qwen3-coder-480b-a35b-instruct | 262,144 | Coding, Text; reasoning, long context, code reasoning, agentic coding | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
qwen3-coder-flash | 1,000,000 | Coding, Text; reasoning, long context, code reasoning, code completion, agentic coding | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
qwen3-coder-plus | 1,000,000 | Coding, Text; reasoning, long context, code reasoning, code completion, agentic coding | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
qwen3-max | 262,144 | Audio, Coding, Text, Video, Vision; reasoning, long context, translation, creative writing, spatial vision, document vision, visual coding, speech-to-speech, omni/video understanding, long video, temporal video, code reasoning, code completion, agentic coding | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
qwen3-vl-flash | 131,072 | Coding, Text, Video, Vision; reasoning, long context, translation, creative writing, spatial vision, document vision, omni/video understanding, long video, temporal video, code reasoning, code completion, agentic coding | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
qwen3-vl-flash-2025-10-15 | 131,072 | Coding, Text, Video, Vision; reasoning, long context, translation, spatial vision, document vision, omni/video understanding, long video, temporal video, code reasoning, code completion, agentic coding | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
qwen3-vl-plus | 131,072 | Coding, Text, Video, Vision; reasoning, long context, translation, creative writing, spatial vision, document vision, omni/video understanding, long video, temporal video, code reasoning, code completion, agentic coding | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
qwen3.6-plus | 1,000,000 | Coding, Text, Video, Vision; reasoning, long context, translation, creative writing, spatial vision, document vision, visual coding, omni/video understanding, long video, temporal video, code reasoning, agentic coding | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
qwen3.6-plus-2026-04-02 | 262,144 | Coding, Text, Vision; code reasoning, long context, reasoning, vision | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
qwen3.7-flash | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取、缓存写入 |
qwen3.7-flash-2026-07-15 | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取、缓存写入 |
qwen3.7-max | 1,000,000 | Coding, Text; reasoning, long context, translation, creative writing, code reasoning, agentic coding | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取、缓存写入 |
qwen3.7-max-2026-06-08 | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
qwen3.7-plus | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取、缓存写入 |
qwen3.8-max | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取、缓存写入 |
wan2.7-i2v | — | Video | /v1/video/generations | 按输出秒，分辨率分档 |
wan2.7-image | 不适用 | Image, Text, Vision; reasoning, vision, image generation, image editing, text in images, image consistency | /v1/images/generations, /v1/images/edits | 按次请求 |
wan2.7-image-pro | 不适用 | Image, Text, Video, Vision; reasoning, long context, vision, image generation, image editing, text in images, image consistency, image-to-video | /v1/images/generations, /v1/images/edits | 按次请求 |
wan2.7-r2v | — | Video | /v1/video/generations | 按输出秒，分辨率分档 |
wan2.7-t2v | — | Video | /v1/video/generations | 按输出秒，分辨率分档 |
Moonshot
| Model ID | 上下文 | 能力 | Endpoint(s) | 计费方式 |
|---|---|---|---|---|
kimi-k2-thinking | 256,000 | Coding, Text; reasoning, long context, code reasoning, agentic coding | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
kimi-k2.5 | 262,144 | Coding, Text, Video, Vision; reasoning, long context, spatial vision, document vision, visual coding, omni/video understanding, long video, code reasoning, agentic coding | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
kimi-k2.6 | 128,000 | Text; long context, reasoning | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
kimi-k2.7-code | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取、缓存写入 |
kimi-k3 | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
MiniMax
| Model ID | 上下文 | 能力 | Endpoint(s) | 计费方式 |
|---|---|---|---|---|
MiniMax-M2.5 | 262,144 | Coding, Text; reasoning, long context, creative writing, code reasoning, agentic coding | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
MiniMax-M2.7 | — | Text | /v1/chat/completions, /v1/responses | 按 1M Token：输入、输出、缓存读取、缓存写入 |
MiniMax-M2.7-highspeed | — | Text | /v1/chat/completions, /v1/responses | 按 1M Token：输入、输出、缓存读取、缓存写入 |
MiniMax-M3 | 1,000,000 | Coding, Text, Video, Vision; reasoning, long context, code reasoning, agentic coding, vision, long video | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
Zhipu GLM
| Model ID | 上下文 | 能力 | Endpoint(s) | 计费方式 |
|---|---|---|---|---|
glm-5 | 128,000 | Coding, Text; reasoning, long context, code reasoning, code completion, agentic coding | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
glm-5.1 | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
glm-5.2 | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
glm-5.3 | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
glm-5.3-flash | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
glm-5v-turbo | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
ByteDance
| Model ID | 上下文 | 能力 | Endpoint(s) | 计费方式 |
|---|---|---|---|---|
dreamina-seedance-2-0-260128 | — | Video | /v1/video/generations | 按 1M Token，分辨率分档 |
dreamina-seedance-2-0-fast-260128 | — | Video | /v1/video/generations | 按 1M Token，分辨率分档 |
dreamina-seedance-2-5 | — | Video | /v1/video/generations | 按 1M Token，分辨率分档 |
seed-1-6-250915 | 262,144 | Text, Video, Vision; reasoning, long context, creative writing, vision, omni/video understanding | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
seed-1-6-flash-250715 | 262,144 | Text, Video, Vision; reasoning, long context, spatial vision, omni/video understanding, temporal video | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
seed-1-8-251228 | 262,144 | Coding, Text, Video, Vision; reasoning, long context, creative writing, spatial vision, document vision, visual coding, long video, temporal video, code reasoning, code completion, agentic coding | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
seed-2-0-lite-260228 | 262,144 | Coding, Text, Video, Vision; reasoning, long context, creative writing, spatial vision, document vision, omni/video understanding, long video, temporal video, code reasoning, code completion, agentic coding | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
seed-2-0-mini-260215 | 262,144 | Coding, Text, Video, Vision; reasoning, long context, spatial vision, document vision, omni/video understanding, temporal video, code reasoning, code completion, agentic coding | /v1/chat/completions | 按 1M Token：输入、输出、缓存读取 |
seed-2-0-pro-260328 | 262,144 | Coding, Text, Video, Vision; reasoning, long context, creative writing, spatial vision, document vision, visual coding, omni/video understanding, long video, temporal video, code reasoning, code completion, agentic coding | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
seedream-5-0-260128 | 262,144 | Image, Vision; image editing, image generation, text in images, vision | /v1/images/generations | 按次请求 |
Tencent
| Model ID | 上下文 | 能力 | Endpoint(s) | 计费方式 |
|---|---|---|---|---|
hy3 | — | Text | /v1/chat/completions, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
hy4-preview | — | Text | /v1/chat/completions, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
Xiaomi
| Model ID | 上下文 | 能力 | Endpoint(s) | 计费方式 |
|---|---|---|---|---|
mimo-v2.6-flash | — | Text | /v1/chat/completions | 按 1M Token：输入、输出、缓存读取 |
mimo-v2.6-pro | — | Text | /v1/chat/completions | 按 1M Token：输入、输出、缓存读取 |
mimo-v2.6-pro-ultraspeed | — | Text | /v1/chat/completions | 按 1M Token：输入、输出、缓存读取 |
StepFun
| Model ID | 上下文 | 能力 | Endpoint(s) | 计费方式 |
|---|---|---|---|---|
step-3.5-flash | — | Text | /v1/chat/completions, /v1/messages | 按 1M Token：输入、输出、缓存读取 |
step-3.5-flash-2603 | — | Text | /v1/chat/completions, /v1/messages | 按 1M Token：输入、输出、缓存读取 |
step-3.7-flash | — | Text | /v1/chat/completions, /v1/messages, /v1/responses | 按 1M Token：输入、输出、缓存读取 |
HappyHorse
| Model ID | 上下文 | 能力 | Endpoint(s) | 计费方式 |
|---|---|---|---|---|
happyhorse-1.1-i2v | — | Video | /v1/video/generations | 按输出秒，分辨率分档 |
happyhorse-1.1-r2v | — | Video | /v1/video/generations | 按输出秒，分辨率分档 |
happyhorse-1.1-t2v | — | Video | /v1/video/generations | 按输出秒，分辨率分档 |
Jina
| Model ID | 上下文 | 能力 | Endpoint(s) | 计费方式 |
|---|---|---|---|---|
jina-embeddings-v3 | — | Embeddings | /v1/embeddings, /v1/classify | 按 1M Token：输入（按上游报告的 token 总量） |
jina-embeddings-v5-text-small | — | Embeddings | /v1/embeddings | 按 1M Token：输入（按上游报告的 token 总量） |
jina-reranker-v3 | — | Rerank | /v1/rerank | 按 1M Token：输入（按上游报告的 token 总量） |
TypeSafe
| Model ID | 上下文 | 能力 | Endpoint(s) | 计费方式 |
|---|---|---|---|---|
jev-latest | — | Text | /v1/systemone | 按 1M Token：输入 |
如何选择模型
| 需求 | 可优先尝试 | 原因 |
|---|---|---|
| 前沿文本 + 视觉 | gpt-5.5, claude-opus-4-8, gpt-5.4 | 推理能力强，多模态和代码能力覆盖广。 |
| Agent 编程 | gpt-5.3-codex, claude-opus-4-8, qwen3-coder-plus, MiniMax-M3 | 具备代码、长上下文和 Agent 子能力。 |
| 低成本高频文本任务 | qwen-flash, deepseek-v4-flash, qwen-mt-flash | 常规任务的输入/输出价格较低。 |
| 长上下文中文或双语任务 | qwen3.6-plus, qwen3.7-max, MiniMax-M3 | 提供 1M token 上下文选项，并具备较强中文能力。 |
| 视觉/文档任务 | qwen3-vl-plus, claude-opus-4-8, gpt-5.4 | 带有视觉、文档和空间理解能力标签。 |
| 图像生成 | gpt-image-2, seedream-5-0-260128, wan2.7-image-pro | 支持图像生成和图像编辑，并按请求计费。 |
Agent 使用注意事项
- 不要编造 AIsa 模型 ID。请使用表格中的精确
model字符串。 - 不要假设某个模型支持其上游模型族的所有模态。请以这里列出的能力标签为准，或查看实时模型页面。
- 如果某个模型出现在 aisa.one/models 但未出现在静态表格中，可能是定价 API 已在运行时启用该模型；请优先参考实时目录。
- 本页不列出价格。当前价格见 aisa.one/models；最终计费金额以 AIsa Usage Logs 为准，并可能包含工作区级别的定价规则。
- 上下文列显示
—表示当前元数据导出没有公布该模型的上下文窗口。不要假设一个默认值，请查看实时模型目录。