Usage Guides | UnifyLLM Docs

Service Infrastructure and Network

UnifyLLM's core API service is hosted in Los Angeles, United States. The production cluster consists of three servers, each with a 20-core CPU and 64 GB of memory, providing 60 CPU cores and 192 GB of memory in total, together with 10 Gbps of dedicated network bandwidth.

International dedicated and optimized routes are used for cross-region API traffic, providing strong connectivity from Asia-Pacific, Europe, and North America. Actual latency can still vary with the user's location, local ISP, upstream model load, context length, and model generation speed.

Test representative production requests before scaling traffic. Material node maintenance or infrastructure changes will be announced through the console and the official Telegram group.

Supported Countries and Regions

UnifyLLM currently focuses on serving users in the following major countries and regions:

Asia-Pacific: Japan, South Korea, Singapore, Malaysia, Thailand, Vietnam, the Philippines, Indonesia, India, Australia, and New Zealand.
North America: the United States, Canada, and Mexico.
Europe: the United Kingdom, Ireland, France, Germany, the Netherlands, Belgium, Luxembourg, Switzerland, Austria, Italy, Spain, Portugal, Sweden, Norway, Denmark, Finland, Poland, and Czechia.
Latin America: Brazil, Argentina, Chile, Colombia, and Peru.
Middle East and Africa: the United Arab Emirates, Saudi Arabia, Qatar, Israel, Türkiye, South Africa, Egypt, and Kenya.

Use the Console

Overview and Dashboard: review current balance, total usage, request count, model usage, token consumption, cost trends, and recent activity.
API Keys: create dedicated keys with a group, expiration time, quota, model restrictions, IP allowlist, and optional cross-group retry settings.
Usage Logs: inspect request time, key, model, group, input and output tokens, cache usage, cost, latency, stream status, Request ID, and error details. Task Logs cover asynchronous image and video jobs.
Wallet: add funds, redeem codes, review balance and billing history, and manage affiliate rewards when the feature is enabled.
Balance Alerts: set a quota warning threshold and receive notifications by email, Webhook, Bark, or Gotify.
Playground: test a model, prompt, and API Key before integrating the request into an application.

Credits, Currency, and Refunds

UnifyLLM uses ⚡️ as its platform credit unit. The current reference conversion is:

CNY, HKD, and JPY values are approximate and may move slightly with exchange rates. Actual charges are determined by the rate displayed on the pricing page and in Usage Logs.

Users may request a refund of the unused portion of credits they actually purchased. UnifyLLM does not charge a platform refund fee. Consumed credits, trial credits, gifted credits, promotional rewards, and affiliate earnings are not cash-refundable. The refundable amount is limited to the user's unused paid balance, and the original payment provider's rules may still apply.

⚡️1 = US$0.147
⚡️1 ≈ CNY ¥1
⚡️1 ≈ HKD HK$1.15
⚡️1 ≈ JPY ¥23

Choose a Model Group

UnifyLLM groups models by source, service stability, supply, and price. The long-term groups are default, standard, and prime. Other groups are generally temporary and depend on market supply.

A group multiplier is a billing factor rather than a measure of model intelligence or speed. For the same model base rate, the multiplier affects the final group price. Discount labels are comparative estimates shown by the console; the final charge for each model is determined by the live pricing page and Usage Logs.

default — 1x Multiplier, Approximately 85% Discount: sources are more varied, so model quality, latency, and availability may fluctuate. Pricing is commonly around 15% of official rates. It is best for study, testing, batch work, and non-critical workloads. Test with representative requests before increasing usage, and do not rely on it as the sole route for a critical production system.
standard — 4x Multiplier, Approximately 43% Below prime: sources are generally official APIs or provide near-official model performance. Prices remain lower than official rates, but models may occasionally become unavailable when market supply is constrained. UnifyLLM uses load balancing to maximize continuity. This is the recommended group for most production workloads.
prime — 7x Multiplier, Stability-Focused Reference Tier: prices are close to, and may sometimes be slightly higher than, official rates, so no fixed discount is promised. More resources are committed to maintaining supply and service stability. For production workloads where reliability matters more than price, prime is the preferred choice.
Temporary Groups — Dynamic Multiplier and Discount: groups such as codex, kiro, cli, and ccmax are often named after their supply source. They can offer even better value than default, but their multiplier, discount, availability, and duration can change with upstream supply or risk-control conditions. Use them for replaceable, retryable, or cost-sensitive work and maintain a fallback for critical traffic.

Use Group-Prefixed Models with default

The default group also exposes selected popular models from other groups by adding the target group as a model-name prefix. This lets a single default-group API Key access more than one service tier without creating a separate key for each group.

For example, the following identifier means that a default-group key calls gpt-5.5 through the prime group:

1prime/gpt-5.5

Acceptable Use and Compliance

UnifyLLM may only be used for lawful and compliant personal or business purposes. Users must not use the service to violate applicable laws, public order, provider policies, or the UnifyLLM User Agreement.

When prohibited use is discovered or supported by reasonable evidence, UnifyLLM may immediately suspend service and permanently ban the account. The remaining balance will not be refunded. Where permitted or required by law, evidence may be preserved and reported or transferred to the competent judicial, law-enforcement, or regulatory authority.

Users are responsible for their accounts, API Keys, and all requests made with them. To the maximum extent permitted by applicable law, UnifyLLM is not responsible for liabilities, losses, or third-party claims caused by a user's unlawful or non-compliant use. Read the complete User Agreement before using the service.

Do not conduct fraud, cyberattacks, malicious automation, or other unlawful activity.
Do not generate, distribute, or assist with illegal content.
Do not infringe privacy, intellectual property, or other lawful rights.
Do not bypass safeguards, abuse platform resources, or disrupt normal service.
Do not use the service in a way that may cause harm to individuals, organizations, or the public.

Data Handling and Privacy

UnifyLLM does not collect or persist API prompts, files, model inputs, or model outputs. Request content is processed transiently by the gateway only as needed to forward the request to the selected model provider.

Request content is not used for advertising, profiling, or model training, and UnifyLLM does not sell, rent, or trade user data.

Only the usage metadata required for billing, service statistics, and troubleshooting is retained:

Requests are still delivered to the selected upstream model provider and are therefore also subject to that provider's data-handling policy. Avoid sending passwords, payment details, credentials, or unnecessary highly sensitive information through any third-party model API. See the Privacy Policy for more information.

Request time and Request ID.
API Key identifier, model, and group.
Input, output, and cache token usage.
Actual cost, response time, status, and necessary error information.

API Documentation and Provider References

The UnifyLLM API Reference contains commonly used endpoints, parameters, and examples. Because space is limited and provider capabilities evolve, it does not list every supported parameter or scenario. Consult the original provider documentation when using advanced or provider-specific features.

Announcements, Community, and Support

Model launches and removals, price changes, maintenance, group availability, and other service changes are announced through one or both of the following channels:

Production users should follow both channels. For support, contact @sunsky89757 on Telegram. For API issues, include your user ID, request time, model, group, Request ID, and error message, but never send a complete API Key.

Trial Credits and Enterprise Support

New users can join the official Telegram group and contact @sunsky89757 with their user ID to receive ⚡️3 in trial credit.

Independent founders, companies, and teams can submit a request through the contact page. Approved projects may receive up to US$1,000 in support credits, depending on their use case, expected usage, and review result.

Customers with substantial, predictable long-term usage can request dedicated routes, more competitive pricing, contracts, and invoices tailored to their business needs. Contact @sunsky89757 or email [email protected] to discuss an enterprise plan.

Partners and Model Requests

Partners are welcome to resell the API through systems such as new-api, integrate UnifyLLM into their own products, or introduce the platform to their customers and communities.

Depending on projected volume and the cooperation model, UnifyLLM can offer partner pricing or a long-term recurring commission arrangement. There is no agency or consultation fee; pricing, settlement, responsibilities, and compliance requirements are agreed for each partnership.

If a model you need is not currently listed, contact @sunsky89757 or [email protected]. Models with clear and sustainable demand can be evaluated for onboarding.