Byesu API

English

简体中文
Русский
日本語
Français
Tiếng Việt

English

简体中文
Русский
日本語
Français
Tiếng Việt

Appearance

byesu vs LiteLLM: Hosted Gateway or Self-Hosted Proxy ​

Short version: these are different kinds of tool. LiteLLM is open-source software: a Python SDK and a proxy server that you run yourself in front of provider accounts that you own. byesu is a hosted AI API gateway: you get one key, byesu is the provider-facing side, and there is nothing to deploy. Pick LiteLLM when your team wants to self-host, keep its own provider keys and contracts, and control routing and budgets. Pick a hosted gateway when you want one key for Claude, GPT, Gemini and Grok without running infrastructure. The two also work together, because LiteLLM can call any OpenAI-compatible endpoint.

Last verified: 2026-11-14. Every statement about LiteLLM below comes from docs.litellm.ai or the LiteLLM GitHub repository, read on that date and listed under Sources. Anything those pages did not state is left out. byesu publishes this page.

At a Glance ​

DimensionbyesuLiteLLM
What it isHosted AI API gatewayOpen-source Python SDK and self-hosted proxy server (AI Gateway)
Who runs itbyesuYou: Docker, Helm or Terraform
Provider keysNot needed; one byesu keyYour own keys for each provider
API formatsOpenAI-compatible /v1/chat/completions and /v1/responses; Anthropic-native /v1/messages/chat/completions, /responses, /messages, /embeddings and more
BillingPay-as-you-go by token usage from a prepaid balance; live price in the consoleYou pay each provider directly; the open-source software is free, enterprise features need a license key
Spend controlPer-request usage in the console; quota per tokenVirtual keys with budgets and rate limits per key, team and user (needs a database)
RoutingHandled by the gatewayYou configure fallbacks, retries and load balancing
Data pathRequests pass through byesuRuns in your own infrastructure
Public status dataStatus page, weekly benchmarksNot applicable; you monitor your own deployment

What LiteLLM Is ​

LiteLLM's documentation describes two ways to use it:

  • Python SDK — a library that calls 100+ LLMs through one interface and returns responses in the OpenAI Chat Completions format, whichever provider served them.
  • Proxy server (AI Gateway) — described as a self-hosted gateway for platform teams managing LLM access across an organization. It gives each user, team and project a virtual key with its own models, budgets and rate limits, records spend by key, team, user and tag, handles load balancing, fallbacks and retries, and includes an Admin UI.

The proxy exposes /chat/completions, /responses and an Anthropic-format /v1/messages endpoint, among others. Deployment options are Docker, Helm and Terraform. The quick start notes that virtual keys require a database and that budgets are not enforced without one; the Docker Compose quick start runs the proxy alongside PostgreSQL. The project is MIT-licensed, and an enterprise license key on the same image enables features such as SSO, audit logs and team admin roles.

What byesu Is ​

byesu is a hosted AI API gateway. You create a key in the console, set a base URL and choose a model:

  • OpenAI-compatible clients use https://byesu.com/v1 (/v1/chat/completions and /v1/responses).
  • Anthropic-format clients such as Claude Code use https://byesu.com (/v1/messages).

One key covers Claude, GPT, Gemini, Grok and other models. Billing is pay-as-you-go by token usage with no subscription; the live price of each model is shown in the console after sign-in, and the console records usage per request. There are setup guides for Claude Code, Codex CLI, Cherry Studio and other clients. Availability and latency are published on the status page and the weekly benchmarks page.

What byesu does not offer: you cannot bring your own provider keys, there is no self-hosted version, and it does not provide the organization features LiteLLM documents, such as per-team budgets, SSO or guardrails.

Where LiteLLM Is the Right Choice ​

  • You need your own provider contracts. Negotiated pricing, committed spend, cloud credits or data agreements only apply when requests use your own accounts. LiteLLM is built around that.
  • Prompts must stay in your environment. LiteLLM's enterprise page describes it as self-hosted in your VPC, so prompts and responses do not leave your environment on the way to the provider.
  • You manage access for many people. Virtual keys, budgets and rate limits per team or user, spend tracking by tag and an Admin UI are the core of the proxy.
  • You want to define routing yourself. Fallback order, retries and load balancing across deployments are yours to configure.
  • You call LLMs from Python and only want one interface. The SDK alone does that with no server at all.

Where a Hosted Gateway Is the Right Choice ​

  • You do not want to run infrastructure. No proxy, database, upgrades or monitoring to own.
  • You want one key and one balance instead of an account, a billing setup and rate limits with each provider.
  • You are an individual or a small team using coding tools such as Claude Code or Codex CLI and want to switch between model families with the same key.
  • You want to start in minutes: create a key, set the base URL, pick a model (Quick Start).

The trade-off is the one listed above: with a hosted gateway your requests pass through a third party and you use its catalog and its prices instead of your own provider agreements.

Using Them Together ​

The two are not exclusive. LiteLLM's OpenAI-compatible provider lets you point a model at any OpenAI-compatible endpoint by adding the openai/ prefix to the model name and setting api_base. That makes byesu one upstream in a LiteLLM proxy, next to your own provider keys:

yaml
model_list:
  - model_name: claude-opus-5-5
    litellm_params:
      model: openai/claude-opus-5-5        # openai/ prefix = OpenAI-compatible route
      api_base: https://byesu.com/v1
      api_key: os.environ/BYESU_API_KEY
bash
litellm --config config.yaml

This syntax follows LiteLLM's OpenAI-compatible endpoints and proxy config pages (checked 2026-11-14): api_base should end in /v1 with nothing after it, and os.environ/NAME reads the value from an environment variable. The model name after openai/ must match the byesu console list exactly. A full walkthrough, including the Python SDK, is in Using byesu with LiteLLM.

A common arrangement: a team runs LiteLLM for virtual keys and budgets, uses its own provider keys for the models it has contracts for, and adds a hosted gateway as another upstream for the rest.

Which One Should You Pick? ​

Choose LiteLLM if you:

  • Want to self-host and keep prompts inside your own infrastructure
  • Have, or want, direct accounts and contracts with providers
  • Need per-team keys, budgets, rate limits and spend reports
  • Have someone to operate a proxy and its database

Choose byesu if you:

  • Want one key for Claude, GPT, Gemini and Grok with nothing to deploy
  • Prefer pay-as-you-go from a single balance over separate provider accounts
  • Mainly use clients such as Claude Code, Codex CLI or Cherry Studio
  • Are fine with requests passing through a hosted gateway

Use both if you already run LiteLLM and want to add models through one more OpenAI-compatible upstream.

FAQ ​

What is the difference between byesu and LiteLLM? ​

LiteLLM is open-source software, a Python SDK and a self-hosted proxy, that you run in front of provider accounts you own. byesu is a hosted AI API gateway: you use one byesu key, do not bring provider keys and do not deploy anything. LiteLLM gives more control and takes operations work; byesu removes the operations work and gives less control.

Is LiteLLM free? ​

The LiteLLM SDK and proxy are open source under the MIT license, and you pay your model providers directly for usage. Per the LiteLLM enterprise page (checked 2026-11-14), features such as SSO, audit logs and team admin roles are enabled by an enterprise license key. You also carry the cost of hosting the proxy and its database.

When should I choose LiteLLM over a hosted gateway? ​

Choose LiteLLM when your team wants to self-host, keep its own provider keys and contracts, keep prompts inside its own infrastructure, and control routing, budgets and rate limits per team or user. It needs someone to deploy and operate the proxy, and a Postgres database for virtual keys and budget enforcement.

Can I use byesu with LiteLLM? ​

Yes. LiteLLM can call any OpenAI-compatible endpoint: add the openai/ prefix to the model name, set api_base to https://byesu.com/v1 and pass your byesu key as api_key. This works in both the Python SDK and the proxy config.yaml, so byesu can be one upstream next to your own provider keys.

Does byesu support bring-your-own-key or self-hosting? ​

No. byesu is a hosted gateway and is the provider-facing side, so you do not add your own provider keys and there is no self-hosted version. If you need either, LiteLLM or another self-hosted gateway is the better fit.

Sources ​

All pages read on 2026-11-14.

Got a problem? Contact support or join our group.

© Byesu API · 半页酥