Pricing & Billing

Credits

Yunxin bills with a single, provider-agnostic unit: credits.

You fund an account with credits in the Dashboard, and each request deducts credits based on the model used and the tokens consumed. Because everything is denominated in credits, one balance covers every provider and modality.

Per-model token pricing

Each model carries its own prices, stored in credits per 1,000,000 tokens. The full price object is returned by the Models API under each model's pricing field:

# Inspect a model's pricing object
curl https://api.yuhuanstudio.com/v1/models/model-id \
  -H "Authorization: Bearer $YUNXIN_API_KEY"

How a request's cost is computed

For a text request, the cost in credits is approximately:

cost ≈ (input_tokens − cached_tokens) · input
     + cached_tokens               · cache_read
     + output_tokens               · output
     + reasoning_tokens            · output

(scaled by the per-1,000,000-token rates above). Other modalities — image, audio, video, music, 3D — are priced per generated unit; consult the model's pricing and the Dashboard.

Every request's token counts, computed cost, latency, routing, and fallback are recorded in your usage history. See Usage Monitoring Tools to read them back.

Discounts

• Batch API runs at roughly a 50% discount — see Batch API.
• When a request is rerouted, a fallback discount may apply; the amount is reported in the X-Fallback-Discount response header. See Fallback & Routing.

Insufficient balance

If your balance can't cover a request, the API returns 402 with code insufficient_balance:

{
  "error": {
    "code": "insufficient_balance",
    "message": "Your credit balance is too low to complete this request."
  },
  "timestamp": 1709251200
}

Top up in the Dashboard to resume. Monitor your balance proactively via the usage endpoints so you can alert before you hit this. See Error Handling.

Tiers

Your account tier sets your request rate limit, request priority, and which models you can reach.

See Rate Limits for how these limits are enforced and the headers that report them.

Request priority

Your tier also determines a request priority level, surfaced in the X-Request-Priority response header: 1 (low), 2 (normal), 3 (high), 4 (critical). Higher-tier traffic is favored under load.

Model access

Some models require a minimum tier. If your account tier is below a model's requirement, the API returns 403 with code model_tier_restricted. Upgrade your tier in the Dashboard to gain access, and check GET /v1/models to see which models your account can use.

Cost optimization

1. Right-size the model — use a smaller, cheaper model for simple tasks; reserve capable models for hard reasoning. Compare prices via the Models API.
2. Set max_tokens to cap output length.
3. Use the Batch API for non-urgent bulk work (~50% off).
4. Cache stable prompt prefixes so repeated input bills at the cache_read rate.
5. Monitor usage via the Dashboard or Usage Monitoring Tools to spot optimization opportunities.

See Best Practices for the full reliability and cost playbook.