Verified Chinese AI access for global teams

Chinese AI models.
One familiar API.

Use the OpenAI client you already know to test Chinese models without rebuilding your integration. RelyRoute is in private beta with one verified route today: DeepSeek V4 Flash.

See what is live

Compatible: OpenAI SDK
External call verified: DeepSeek V4 Flash
Launch state: Closed beta

Route tracerr_84f2a9

Live

RequestSend an OpenAI-format request

Policy$0.01 capped test key

RouteDeepSeek V4 Flash

✓

Latency1.02s

Response200 OK

MeteringLogged

PolicyBudget capped

HealthFlash route verified

What is live today—and what is not

One endpointRequest-level logsCapped test keysNo public payments yet

Verified model catalog

Know what is live before you write code.

DeepSeek V4 Flash is the only customer route verified end to end today. Every other model stays marked as review or planned until its pricing, terms, limits, and reliability are checked.

DS Live

DeepSeek V4 Flash

DeepSeek

Verified from an external client: HTTP 200, visible response, and token usage recorded.

FastMultilingual

DS Pricing review

DeepSeek V4 Pro

DeepSeek

The channel exists, but customers cannot use it until pricing and limits are validated.

ReasoningComplex tasks

QW Planned

Qwen Plus

Alibaba Cloud

Candidate route. Regional access, resale terms, pricing, and reliability still need review.

GeneralMultilingual

KM Planned

Kimi

Moonshot AI

Candidate route. Long-context behavior, API terms, and international availability still need review.

Long contextResearch

GL Planned

GLM

Zhipu AI

Candidate route. Tool use, billing, and international access still need review.

AgentsTools

SD Planned

Seed

ByteDance

Candidate route. Throughput, regional pricing, and data handling still need review.

ThroughputCreative

The request path

One integration for access, limits, and usage—not another black box.

Your app sends a familiar OpenAI-format request. RelyRoute checks the API key and remaining budget, forwards it to the verified upstream model, records tokens and cost, then returns the response.

See the exact request ↗

One compatible endpoint

Keep your OpenAI client and change only the base URL, API key, and model ID.

Request pathYour app → RelyRoute → DeepSeek

Spend controls per key

Set a hard quota, expiration time, and model allowlist before a key reaches a tester.

Your appVerified model

Request-level usage logs

Review model, input tokens, output tokens, status, and charge for every call.

Control every test key

Keys, limits, and usage in one place.

Create a separate key for each tester or project, cap its spend, restrict its models, and inspect every request. The console below uses sample data and is not connected to public billing.

OverviewPrivate beta workspace

Product preview · sample data

Spend this month$0.84of $10.00 budget

Requests12,486↗ 18.4%

Tokens8.2MInput + output

Success rate99.97%Healthy

UsageRequests by day

1,5001,0005000

Route mixBy request

100%verified

DeepSeek V4 Flash12,486

Five-minute integration

Keep your OpenAI client. Change the base URL and model ID.

Use the familiar Chat Completions request shape. The api.relyroute.com address below is reserved but not active until the custom domain is configured; beta testers receive the working endpoint directly.

✓ Chat Completions request shape
✓ End-to-end external call verified
✓ Per-key budgets and expiry

curl https://api.relyroute.com/v1/chat/completions \
  -H "Authorization: Bearer $RELYROUTE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-v4-flash",
    "messages": [{"role": "user", "content": "Hello"}]
  }'

Private beta

Test one verified route before we add more.

RelyRoute is not open for self-service signup or payment. The closed beta starts with capped, expiring keys for DeepSeek V4 Flash while we validate billing accuracy, reliability, support, and abuse controls.

Review the beta setup ↗

01No public checkout yet

Payments stay closed until pricing and risk controls are ready.

02Capped beta keys

Each key can be limited by budget, model, and expiry.

03Verified before marked live

A model is marked live only after an external request and usage-log review.

Chinese AI models.One familiar API.