Two months ago my side projects were bleeding money on LLM calls. Not because the projects were expensive to run — because official API pricing punishes exactly the usage pattern hobby projects have: bursty, unpredictable, and never large enough to qualify for volume tiers.
So I built a relay instead of paying it. This post is the numbers, the code, and the parts that break.
Disclosure: it's mine. Everything below is measured from my own account, and yes there's a link at the end.
The numbers that made the decision obvious
Per 1M tokens, input / output, USD — official price first, then mine:
| Model | Official | Here |
|---|---|---|
| Grok 4.7 | 2 / 6 | 0.10 / 0.30 |
| GPT-6 Astra | 10 / 50 | 0.60 / 3.00 |
| GPT-5.6 Sol | 4 / 20 | 0.24 / 1.20 |
| GLM-5.3 | 1.4 / 4.4 | 0.105 / 0.33 |
| Gemini 4 Argon | 4 / 20 | 0.48 / 2.40 |
| DeepSeek V4 Pro | 1.32 / 3.96 | 0.198 / 0.594 |
| Claude Sonnet 5 | 2 / 10 | 0.945 / 4.72 |
All of them answer to the same key and drain the same balance.
The design choice that actually matters: 1:1
Most relays in this corner of the market price in points or apply a multiplier on top of a token cost. You end up doing mental arithmetic to find out what a dollar bought you.
This one is 1:1. Pay $1, hold $1 of balance. Usage is deducted at the model's rate. No top-up bonus, no multiplier, no minimum tier.
That property is worth more than any single discount, because it makes cost predictable: the price table is the bill.
Calling it is a one-line change
from openai import OpenAI
client = OpenAI(
api_key="sk-...",
base_url="https://api.daluobo.cc.cd/v1",
)
resp = client.chat.completions.create(
model="gpt-6.2-astra",
messages=[{"role": "user", "content": "explain etag caching briefly"}],
)
Curl works the same way. So do Cherry Studio, Lobe Chat, Cline and Continue — they all just take a base_url.
Where it breaks, honestly
The /v1 suffix is not optional. Drop it and you get an HTML page back instead of JSON, which surfaces in most clients as a bare request failed.
Model IDs are relay-side names, not vendor names. gpt-6.2-astra may not resolve on the official API, and vice versa. Keep a per-project constant instead of string literals scattered through the codebase.
Streaming is fine, exotic vendor extensions are not. Chat Completions with SSE works. Provider-specific params, proprietary tool formats and some multimodal shapes depend on whichever upstream route you land on that minute.
It is not the official API and does not pretend to be. No SLA. For anything revenue-bearing, keep a second provider configured and a way to switch.
The uncensored models are the sleeper feature
Two models — deepseek-v4.1-flash-uncensored and -fast — run with the safety layer removed, at roughly half of official DeepSeek pricing.
For fiction, roleplay and long-form writing this is a different product, not a discounted one. No preamble, no hedging, no refusal when a scene involves anything commercially licensed.
When you should not bother
If your monthly spend is under a few dollars, the official free tiers beat any relay on convenience, and you skip the trust question entirely.
If your workload is heavily multimodal or depends on provider-specific APIs, a relay adds a translation layer you will eventually fight.
Everyone else: the arithmetic is not close.
Base URL https://api.daluobo.cc.cd/v1. The price list is public at daluobo.cc.cd/pricing, and sign-up drops $0.50 of credit on the account so you can measure it before paying anything.
Top comments (0)