YunWu API Review: Pricing & Comparison

GitHub login, free daily GPT-4o calls

YunWu API’s (yunwu.ai) barrier to entry is close to zero: log in with your GitHub account and immediately get a free daily GPT-4o call allowance — no top-up, no credit card required.

This “free daily allowance” is a genuine, usable call quota, not a demo key limited to a handful of tries — for individual developers with light daily use (tens to a few hundred calls), that may well be enough. Once you exceed the free allowance, you pay as you go at an exchange rate of roughly ¥0.5/USD.

What ¥0.5/USD actually means

The market exchange rate is around ¥7.2/USD; YunWu API’s rate is ¥0.5/USD, equivalent to about 6.9% of official USD pricing:

Model	Official USD price	YunWu API converted price	Savings
Claude Opus 4.8 (input)	$5/M tokens	¥2.5/M	93%
Claude Opus 4.8 (output)	$25/M tokens	¥12.5/M	93%
GPT-5.5 (input)	$5/M tokens	¥2.5/M	93%
DeepSeek V4 Pro	$0.27/M	¥0.14/M	93%

This is one of the most aggressive exchange rates among comparable platforms (TokenRiver is 1¥=$1, about 14%; bltcy is 1-2.5¥/USD, about 14-36%).

A detail worth verifying yourself: with a rate this low, it’s reasonable to wonder whether some models run through official channels. We’d suggest comparing output quality after topping up, or checking whether the model field returned by the API matches the expected version.

500+ model coverage

YunWu API aggregates 500+ models, mainly including:

• OpenAI: GPT-5.5, GPT-4o (including the free daily allowance), the o-series
• Anthropic: Claude Fable 5, Opus 4.8, Sonnet 4.6, Haiku 4.5
• Google: Gemini 3.5 Flash/Pro
• DeepSeek: V4 Pro, V4 Flash
• Multimodal: Suno music generation, some Midjourney endpoints

The “500+” figure includes a large number of version snapshots and community-maintained models — in practice, focus on the mainstream models listed above and avoid depending on edge-case models.

# Integration example (OpenAI format)
from openai import OpenAI

client = OpenAI(
    api_key="your key",
    base_url="https://yunwu.ai/v1"
)

# Free daily GPT-4o calls after GitHub login
response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Help me write a piece of Python code"}]
)

The trade-off between price and stability

YunWu API doesn’t position stability as its core selling point — that’s a positioning choice, not a hidden flaw. Compared with 4SAPI (enterprise-grade SLA) or KoalaAPI (24h refund guarantee), YunWu API’s value proposition is clear: extremely low price, with stability left for you to verify yourself.

• No public SLA: no compensation commitment for service interruptions
• No latency figures: we’d suggest testing in your own real-world use case
• Channels not fully disclosed: whether some models run through official channels needs to be verified in practice

Well suited to dev/test environments and small personal projects, not to production services that need continuous stability.

Good fit scenarios

• Light users for whom a GitHub account is enough: the free daily allowance covers everyday testing at zero cost
• Students and beginners: ¥0.5/USD lowers the cost of exploring AI APIs about as far as it can go
• Cross-model comparison: 500+ models under one account, great for quickly switching between and comparing different models
• Vibe-coding personal projects: no monthly fee, top up in small amounts as needed, no waste

For production-grade daily Claude Code usage, we’d suggest pairing YunWu API with a more stable primary platform, using it as a backup or a low-cost testing entry point instead.

On channel risk: GPTGOD (0.6¥/USD exchange rate) is another aggressively low-priced option, but it runs through reverse-engineered channels and is less stable; YunWu API’s low pricing still carries a somewhat higher share of official channels among comparable platforms. If you prioritize “officially authorized channels with no capability downgrade,” ofox.ai claims all its models run through official Azure/AWS/Anthropic channels — priced higher than YunWu API but with more assured channel credibility.

Information verified 2026-08-11. Defer to the official site for the current free daily GitHub allowance policy; the ¥0.5/USD exchange rate reflects live official pricing, and we’d suggest testing some models’ channel quality yourself.

Related reviews

• HuggingFace Inference API: one-click access to hundreds of thousands of open-source models, a favorite of the ML community
• CloseAI: claims to be Asia’s largest enterprise-grade AI relay platform, natively supports the OpenAI/Claude/Gemini protocols, DPA data agreements available
• Banana AI: lightweight GPU inference, fast custom model deployment, low cold-start time, developer-friendly
• EasyRouter: from Fu Sheng, 15% off site-wide, 40+ model providers, DeepSeek as low as 75% off