AIWave

DeepSeek V4 Pro, on a route you can inspect.

Keep the OpenAI-compatible client, pin deepseek-v4-pro, and save the request record beside the result. The documented limits are a planning reference, not proof that your payload will pass.

Copy the request contract Plan the long-context test

Move the client, then test the boundaries.

The base URL and request shape are the small part. Context, output reserve, timeout, 429 handling, cache shape, and the charge record decide whether the route belongs in production.

Route test checked 2026-08-27
01

Set the route

Use https://aiwave.live/v1 as the base URL and load AIWAVE_API_KEY from the environment.

02

Pin the model ID

Send deepseek-v4-pro and save that exact ID with the result.

03

Run the representative payload

Exercise the context, timeout, 429, cache, and output boundaries your application must handle.

Use the dated AIWave gateway rate.

The snapshot separates input, cache-hit input, and output. The effective request group determines the multiplier applied to these base fields.

AIWave snapshot · 2026-08-27 · listed base rate
AIWave gateway long-context rates checked 2026-08-27, shown as listed base rates
RouteInput / 1MCache hit / 1MOutput / 1M
deepseek-v4-flash$0.638$0.0202884$1.914
deepseek-v4-pro$1.914$0.0637362$5.742
V4 Pro gateway row.
Input $1.914 · cache hit $0.0637362 · output $5.742 per 1M token units. Verify the live rate and effective group before production use.

Copy the route, then add your operating rules.

The examples show the client contract. Add your payload, timeout, bounded retries, request logging, and acceptance check.

Python, Node.js, OpenAI-compatible

Python

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["AIWAVE_API_KEY"],
    base_url="https://aiwave.live/v1",
)

result = client.chat.completions.create(
    model="deepseek-v4-pro",
    messages=[{"role": "user", "content": "Your prompt"}],
)

Node.js

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.AIWAVE_API_KEY,
  baseURL: "https://aiwave.live/v1",
});

const result = await client.chat.completions.create({
  model: "deepseek-v4-pro",
  messages: [{ role: "user", content: "Your prompt" }],
});

Keep the first failure.
For a 429, use bounded backoff and inspect the policy. For overflow, reduce input or reserved output before retrying. Preserve the original request evidence.
Acceptance step

Run one representative payload on this exact model ID.

Keep the answer, status, token fields, effective group, and charge together. Decide from that record, not from the size of the documented window.

Prepare the V4 Pro request Review the dated gateway rate