doubao-seed-2-0-code-preview-260215 — 价格、上下文与性能

doubao-seed-2-0-code-preview-260215 是 Volcengine Ark 的模型,上下文 262.1K,支持 Reasoning, Tools, Files, Vision。输入 $0.44,输出 $2.19,均为每百万 tokens。

概览

厂商Volcengine Ark
上下文262.1K
能力Reasoning, Tools, Files, Vision
API 格式openai, openai-response, openai-response-compact, anthropic, gemini, openai-alpha-search
计费方式p * 0.438356 + cr * 0.0876712 + c * 2.19178) : len <= 128000 ? tier("32k_128k", p * 0.657534 + cr * 0.0876712 + c * 3.28767) : tier("128k_plus", p * 1.31507 + cr * 0.0876712 + c * 6.57534
ByteDance Seed coding model for multimodal software engineering and long-running agents

价格

套餐倍率输入输出缓存输入说明
基准价1×$0.44$2.19$0.09
Max4×$1.75$8.77$0.35Premium — max reasoning & stability.

仅列出该模型已开放的套餐。

性能实测

API 端点https://api.airai.cc/v1
OpenAI 兼容OpenAI-compatible

常见问题

doubao-seed-2-0-code-preview-260215 的价格是多少?
输入 $0.44,输出 $2.19,均为每百万 tokens。
doubao-seed-2-0-code-preview-260215 的上下文窗口有多大?
doubao-seed-2-0-code-preview-260215 的上下文窗口为 262.1K,即单次请求中输入与输出合计的 token 上限。
如何在代码中调用 doubao-seed-2-0-code-preview-260215?
把 base_url 指向 https://api.airai.cc/v1,model 填 doubao-seed-2-0-code-preview-260215 即可,接口与 OpenAI 完全兼容。
doubao-seed-2-0-code-preview-260215 为什么会返回 429?
429 是限流响应。降低并发、加退避重试,或切换到配额更高的套餐即可。

模型对比

Model厂商输入输出上下文
doubao-seed-2-0-code-preview-260215Volcengine Ark$0.44$2.19262.1K
doubao-seed-1-6-251015Volcengine Ark$0.11$1.10256K
doubao-seed-2-0-mini-260428Volcengine Ark$0.03$0.27256K
doubao-seed-2-1-pro-260628Volcengine Ark$0.82$4.11256K
doubao-seed-2-0-lite-260428Volcengine Ark$0.08$0.49256K
doubao-seed-1-6-vision-250815Volcengine Ark$0.11$1.10256K
doubao-seed-2-1-turbo-260628Volcengine Ark$0.41$2.05256K
doubao-seed-evolvingVolcengine Ark$0.82$4.11256K
doubao-seed-2-0-pro-260215Volcengine Ark$0.44$2.19256K
doubao-seed-1-6-flash-250828Volcengine Ark$0.02$0.21256K
doubao-seed-1-8-251228Volcengine Ark$0.11$1.10256K

错误排查

429 是限流响应。降低并发、加退避重试,或切换到配额更高的套餐即可。

常见原因

  • 并发请求过多
  • 当前套餐配额已用完
  • 失败后立即重试没有退避

解决办法

  • 限制并发并用指数退避重试
  • 切换到配额更高的套餐
  • 对重复请求做缓存

如何接入

curl https://api.airai.cc/v1/chat/completions \
  -H "Authorization: Bearer <YOUR_API_KEY>" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "doubao-seed-2-0-code-preview-260215",
    "messages": [{"role": "user", "content": "Hello"}]
  }'
from openai import OpenAI

client = OpenAI(
    api_key="<YOUR_API_KEY>",
    base_url="https://api.airai.cc/v1"
)

response = client.chat.completions.create(
    model="doubao-seed-2-0-code-preview-260215",
    messages=[{"role": "user", "content": "Hello"}]
)
print(response.choices[0].message.content)
import OpenAI from "openai";

const client = new OpenAI({
  apiKey: "<YOUR_API_KEY>",
  baseURL: "https://api.airai.cc/v1"
});

const response = await client.chat.completions.create({
  model: "doubao-seed-2-0-code-preview-260215",
  messages: [{ role: "user", content: "Hello" }]
});
console.log(response.choices[0].message.content);
const OpenAI = require("openai");

const client = new OpenAI({
  apiKey: "<YOUR_API_KEY>",
  baseURL: "https://api.airai.cc/v1"
});

client.chat.completions.create({
  model: "doubao-seed-2-0-code-preview-260215",
  messages: [{ role: "user", content: "Hello" }]
}).then((r) => console.log(r.choices[0].message.content));

将 <YOUR_API_KEY> 替换为令牌设置中的 API Key。

兼容端点

openaiopenai-responseopenai-response-compactanthropicgeminiopenai-alpha-search

Gemini SDK · TypeScript

import { GoogleGenerativeAI } from "@google/generative-ai";

const genAI = new GoogleGenerativeAI("<YOUR_API_KEY>");
const model = genAI.getGenerativeModel({ model: "doubao-seed-2-0-code-preview-260215" });

const result = await model.generateContent(
  "Explain quantum entanglement in one paragraph."
);
console.log(result.response.text());

身份验证

所有请求必须携带 Authorization: Bearer <TOKEN> 请求头;Anthropic 格式端点也接受 x-api-key 请求头。在「令牌」页面生成 API Key,可按模型、分组、IP、速率等维度精细化授权。

支持的参数

参数类型默认值范围说明
temperaturefloat10–2采样温度;越低越稳定
top_pfloat10–1核采样累计概率
max_tokensinteger——响应中最大 token 数
frequency_penaltyfloat0−2–2惩罚高频 token 的重复出现
presence_penaltyfloat0−2–2鼓励引入新话题
stopstring[]null≤ 4最多 4 个停止生成的字符串
seedintegernull—尽量保证可复现的采样种子
ninteger1—生成的候选条数
streambooleanfalse—通过 SSE 流式返回 token
response_formatobject——强制输出 JSON 对象或符合 Schema 的结果
toolsarray——模型可调用的工具 / 函数声明
tool_choicestring | objectauto—工具选择策略或具体工具名
logprobsbooleanfalse—返回每个 token 的对数概率
top_logprobsintegernull0–20每个 token 返回的 top 概率数量
logit_biasmap——按 token 的 logit 偏置映射
userstring——用于风险审计的终端用户标识

速率限制

分组RPMTPMRPD
Eco990398K20K
Test980393K20K

数据更新于: 2026-10-10 08:15

技术支持在线客服
返回顶部