UniAPI Review: Pricing & Comparison
OpenAI/Claude/Gemini three-way coverage + Midjourney/Suno/Kling and other multimodal models, mainland direct connect, channel-discount billing
Last verified: 2026-06-13 · Visit official site →
Two routes, so you don’t have to guess about network conditions
Most relay platforms have a single API endpoint, and access from mainland China is a matter of luck. UniAPI explicitly provides two routes:
- Mainland-optimized route:
hk.uniapi.io— a Hong Kong node specifically optimized for mainland access, no proxy needed - Global route:
api.uniapi.io— an overseas node suited to teams with existing proxy infrastructure or servers abroad
There’s also an easily overlooked plus: status.uniapi.ai is UniAPI’s independent status monitoring page, showing real-time availability of each node and the main models. You can add this URL to your monitoring system and set up alerts, instead of finding out something’s wrong only after users complain.
Tiered discounts: breaking down the billing formula
UniAPI’s billing formula:
Actual cost = token count × official model rate × channel discount × tier discount
Two layers of discount stack together:
| Discount layer | Description | Example |
|---|---|---|
| Channel discount | Same model, different upstream prices | Azure channel ~50% off, OpenAI official channel ~32% off |
| Tier discount | Higher top-up amount/call volume unlocks a bigger extra discount | Up to an additional ~15% off |
A worked example: Claude Opus 4.8’s official price is $5/M input tokens (≈¥36/M). Through the Azure channel (50% off) stacked with the tier discount (15% off), that comes out to roughly ¥15.3/M — about 42.5% of the official direct price.
The exact discount structure and tier thresholds follow the live configuration on the official site — we’d recommend logging into the dashboard to check current pricing before estimating costs.
Three protocols + multimodal, all under one account
Core LLMs (three native protocols):
# Anthropic native (configure directly in Claude Code, full Prompt Caching support)
export ANTHROPIC_BASE_URL="https://hk.uniapi.io"
export ANTHROPIC_API_KEY="your key"
# OpenAI format (GPT/DeepSeek/domestic models)
export OPENAI_BASE_URL="https://hk.uniapi.io/v1"
export OPENAI_API_KEY="your key"
# Gemini native (configure directly in Gemini CLI)
export GEMINI_BASE_URL="https://hk.uniapi.io"
Multimodal extensions (same account, same endpoint):
| Type | Available models |
|---|---|
| Image generation | Midjourney, FLUX, Replicate |
| Music generation | Suno (AI music creation) |
| Video generation | Kling |
Multimodal content is accessed through the UniAPI endpoint, saving you the hassle of separately maintaining a Midjourney Discord account, a Suno account, and so on.
Native Anthropic protocol support means the cache_control field for Prompt Caching doesn’t get stripped by a translation layer — in practice this can save a considerable amount of token cost in scenarios like large codebase analysis.
Model coverage
Claude Fable 5, Opus 4.8 (Anthropic native), GPT-5.5, the o-series, Gemini 3.1 Pro (Gemini native), DeepSeek V4, Qwen3 — the mainstream flagships are all there, with full Suno/Midjourney/Kling multimodal coverage.
Which teams it suits
| Scenario | Is UniAPI a fit? |
|---|---|
| Daily Claude Code usage | Yes (native Anthropic protocol, zero adaptation) |
| Midjourney + LLM combined workflows | Yes (unified multimodal entry point) |
| Need an independent status page for monitoring | Yes (status.uniapi.ai) |
| Chasing the absolute lowest price | Consider TokenRiver (1¥=$1 rate, ~14% of official price) instead |
| Need a DPA enterprise compliance agreement | Contact UniAPI’s sales team directly to confirm |
On pricing transparency, UiUiAPI publishes each model’s discount rate quantitatively on its official site, which is better suited to scenarios where you want to precisely calculate cost before topping up, rather than estimating from stacked channel discounts.
Information verified 2026-06-13. Discount structure and tier thresholds follow live information on the uniapi.ai official site.
Related reviews
- 302.AI: pay-as-you-go with zero monthly fee
- 4SAPI / Starlink 4SAPI: enterprise-grade
- B.AI: full model lineup + USDT crypto payment, privacy-friendly, no real-identity binding required
- PaintBot: OneAPI-panel-driven, 0.5¥/USD rate, standardized interface
Quick facts
| Pricing model | Pay-as-you-go: actual cost = token count × official model rate × channel discount × tier discount, channel discount as low as 50%, plus tier discount up to 85% |
|---|---|
| Model coverage | OpenAI/Claude (native protocol)/Gemini (native protocol) three-way coverage, plus Midjourney/Suno/Kling/Replicate/FLUX and other multimodal models |
| Latency / SLA | Provides a status.uniapi.ai status monitoring page; no specific latency numbers published |
| Mainland direct connect | Mainland direct connect (hk.uniapi.io mainland-optimized route), plus api.uniapi.io overseas route |
| Best for | Enterprise / Developers |
| Referral program | No public affiliate program found. |
Pros
- Unified access to the three major model families (OpenAI/Claude/Gemini, including native protocols), extended with support for Midjourney/Suno/Kling/Replicate and other image/video/audio generation models
- Offers both a hk.uniapi.io mainland-optimized direct-connect route and an api.uniapi.io global route, so mainland users don't need to configure an extra proxy
- Uses a "channel discount × tier discount" billing structure — the more you top up or the higher your tier, the bigger the discount (up to roughly 85% off combined) — and publishes a dedicated status monitoring page (status.uniapi.ai)
Cons
- Public information on specific latency numbers and SLA compensation terms is limited — check the official site/docs directly to confirm
- No public affiliate/referral program found
Compare more AI API relays
See the full comparison board — filter by price tier, model coverage, and mainland direct-connect status.
Back to the comparison board →