CLAUDE · GPT · GPT-IMAGE
Straight to the model.At list price.
Real Claude, native GPT and gpt-image over stable overseas residential IPs. Top up at ¥1 = $1, resold at a 1× multiplier, key in seconds.
Used for Claude Code, AI products, and image generation at volume.
Running in production
- 1.2M+
- Calls served
- 1×
- Price multiplier
- 86%
- Saved on settlement
- Claude
- GPT
- gpt-image
- Claude Code
- Codex
- Residential IP
PRICING · MULTIPLIER
Same model.One exchange rate apart.
Same models, same rate card. The only difference is settlement: paying in dollars means paying the real exchange rate, while topping up here credits ¥1 as $1. That spread is the saving.
gpt-image · per image
| Size | Same USD · at the real rate | hopimg · ¥1 = $1 |
|---|---|---|
| 1024 × 1024 | ¥0.95 | ¥0.13 |
| 2048 × 2048 | ¥1.43 | ¥0.20 |
| 4096 × 4096 | ¥1.90 | ¥0.27 |
Chat · per million tokens
| Model | Same USD · Input / Output | hopimg · Input / Output |
|---|---|---|
| claude-opus-5 | ¥35.50 / ¥177.50 | ¥5.00 / ¥25.00 |
| claude-sonnet-5 | ¥14.20 / ¥71.00 | ¥2.00 / ¥10.00 |
| gpt-5.6-sol | ¥35.50 / ¥213.00 | ¥5.00 / ¥30.00 |
| gpt-5.6-luna | ¥1.42 / ¥8.52 | ¥0.20 / ¥1.20 |
These are the amounts the console actually charges, derived from production billing records. The full model list lives under Models in the console; official Claude Code and Codex connect directly.
≈ 1.40% of list · The "same USD" column converts at $1 ≈ ¥7.1, for comparison only.
Rates verified against production billing records on 2026-08-28. Console billing is authoritative.
THREE USES, ONE GATEWAY
Chat, agents and images,on one account.
- 01Docs
ChatClaude · GPT
Anthropic- and OpenAI-compatible chat endpoints, straight to the real models, streaming included. Existing SDKs only change Base URL.
- 02Docs
AgentsClaude Code · Codex
Point the official CLI at us with one environment variable. Long runs, continuous edits, usage and spend visible live.
- 03Docs
Imagesgpt-image-2
Text-to-image and image-to-image, OpenAI-compatible, native Chinese prompts, results in seconds.
THREE STEPS
A key in 30 seconds.
- 01
Get a key
Sign up and get a key in 30 seconds. Chat, agents and images each get their own group, isolated from one another.
- 02
Point at hopimg
Set Base URL to api.hopimg.com — chat and images share one address, and your existing OpenAI / Anthropic SDK needs no other change.
- 03
Start calling
Chat hits /v1/chat or /v1/messages, images hit /v1/images. Claude Code and Codex connect directly too.
from openai import OpenAI client = OpenAI( base_url="https://api.hopimg.com/v1", api_key="sk-...",) img = client.images.generate( model="gpt-image-2", prompt="霓虹夜景里的柴犬,电影级灯光", size="1024x1024",)DO THE MATH
Same usage.A month apart in cost.
Drag to your own volume. On the left, buying direct at the real exchange rate. On the right, the same models at the same 1× multiplier on hopimg.
Assumptions: the direct column converts at $1 ≈ ¥7.1; chat is estimated on a blended mix with output at 25%. Console billing is authoritative.
Saved per month
¥1,629/ mo
enough for 12,154 more images
Start now.Put AI in your product.
Get a key for free, no card required. Your first call can go through in seconds.