From four to nine providers: one route, a dated ledger, and evidence you can inspect

[chawkian]

AIWave's live catalog now contains 63 models from 9 Chinese AI providers behind one OpenAI-compatible, dollar-denominated API.

The provider count is not the main change. The useful change is that a team can evaluate routes without changing its client contract:

• one base URL and API key;
• exact model IDs instead of family-name guesses;
• dated input, cache-hit and output rates where those rates are exposed;
• an explicit effective request-group multiplier;
• per-model observed success rates with denominators, including weak rows.

The five added provider families are MiniMAX, Qwen/Alibaba, Doubao/ByteDance, StepFun and Xiaomi MiMo. The original four are DeepSeek, GLM/Zhipu, Kimi/Moonshot and ERNIE/Baidu.

The implementation contract stays small:

import os
from openai import OpenAI

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

response = client.chat.completions.create(
    model="step-3.5-flash-2603",
    messages=[{"role": "user", "content": "Return the three main risks as JSON."}],
)

print(response.choices[0].message.content)
print(response.usage)

Before forecasting a bill, pull the live catalog again and keep the rate date beside the model ID. Do not treat an upstream provider price as an AIWave invoice rate, and do not infer the effective request group from a top-up alone—verify it in the Console.

The full launch note includes the dated StepFun rate rows, a worked 10M-input/2M-cache/1M-output cost ledger and a five-test acceptance plan:

https://aiwave.live/blog/aiwave-9-chinese-ai-providers-2026.html

Current references:

• Catalog: https://aiwave.live/models/
• Pricing ledger: https://aiwave.live/pricing
• Observed route evidence: https://aiwave.live/trust
• Integration: https://aiwave.live/docs/

What would make this more useful in an SDK: a versioned catalog fixture, a pre-dispatch cost estimate, or an endpoint that explains the chosen route?