LiteLLM and OpenRouter both simplify access to many models, but they optimize for different teams. LiteLLM is for control and self-hosting; OpenRouter is for managed convenience. Production agents may still need QVeris for live data and tools.
LiteLLM 和 OpenRouter 都能简化多模型接入，但服务的团队不同。LiteLLM 偏控制权和自托管，OpenRouter 偏托管便利。生产级 Agent 仍可能需要 QVeris 来调用实时数据和工具。
Compare self-hosted LiteLLM with managed OpenRouter.
比较自托管 LiteLLM 与托管式 OpenRouter，并将外部能力发现保持为独立层。
| Question问题 | Choose LiteLLM选择 LiteLLM | Choose OpenRouter选择 OpenRouter | Add QVeris加入 QVeris |
|---|---|---|---|
| Primary need核心需求 | Run your own OpenAI-compatible proxy.自行运行兼容 OpenAI 的代理服务。 | Reach many models quickly with one managed API.通过一个托管 API 快速接入多种模型。 | Let agents call verified data, tools, and external services.让 Agent 调用经过验证的数据、工具和外部服务。 |
| Hosting托管方式 | Self-hosted or enterprise managed options.支持自托管或企业托管方案。 | Managed multi-provider model gateway.托管式多提供商模型网关。 | Managed capability routing network.托管式能力路由网络。 |
| Best buyer最适合团队 | Platform teams, infra teams, regulated teams.平台团队、基础设施团队和受监管团队。 | Developers, startups, fast-moving product teams.开发者、初创公司和快速迭代的产品团队。 | Teams building production agents with real-world tasks.构建生产级 Agent 并执行真实任务的团队。 |
| Core trade-off核心取舍 | More control, more operational responsibility.控制力更强，同时承担更多运维责任。 | Less ops, less direct infrastructure control.运维更少，但对基础设施的直接控制也更少。 | Not a model router; complements your model gateway.它不是模型路由器，而是模型网关的能力补充层。 |
Practical answer: choose LiteLLM if you need control, OpenRouter if you need speed, and QVeris when the agent needs something useful to do after the model responds.
实用答案：需要控制权选 LiteLLM，需要快速接入选 OpenRouter；当 Agent 在模型返回后还要调用真实数据和工具时，加上 QVeris。
Both tools sit in the model layer. They normalize model providers behind a developer-friendly API, but they make different bets on ownership, operations, and product experience.
两者都属于模型层。它们把不同模型供应商统一到更友好的 API 后面，但在所有权、运维和产品体验上的取舍不同。
| Dimension维度 | LiteLLM | OpenRouter |
|---|---|---|
| Positioning定位 | Open-source LLM proxy and gateway.开源 LLM 代理与网关。 | Managed model aggregator and router.托管式模型聚合与路由平台。 |
| Setup部署方式 | Deploy and operate the proxy yourself, unless using an enterprise option.除非采用企业方案，否则需要自行部署和运维代理。 | Create an account, get a key, and call models through one endpoint.注册账户、获取密钥后，即可通过统一端点调用模型。 |
| Control控制能力 | High control over routing, keys, logs, budgets, and provider config.对路由、密钥、日志、预算和提供商配置拥有较高控制权。 | Lower operational burden, but less control over the managed routing layer.运维负担较低，但对托管路由层的控制也更少。 |
| Observability可观察性 | Can be configured around your own proxy and logging stack.可围绕团队自有代理和日志体系进行配置。 | Managed usage visibility and provider-level model access.提供托管式用量可见性和提供商级模型访问。 |
| Pricing model定价模式 | Open-source software cost plus infrastructure and operations.开源软件本身的成本，加上基础设施和运维成本。 | Pay for model usage through a managed provider interface.通过托管式提供商接口按模型用量付费。 |
| Best fit最佳适用场景 | Teams that want infrastructure ownership and policy control.希望拥有基础设施并自行控制策略的团队。 | Teams that want the fastest path to many models.适合希望最快接入多种模型的团队。 |
When LiteLLM is the better choice
什么时候 LiteLLM 更适合
LiteLLM is usually the better fit when your team wants to own the gateway layer and keep model traffic inside your own infrastructure boundary.
当团队希望自己掌控网关层，并让模型流量留在自己的基础设施边界内时，LiteLLM 通常更适合。
When OpenRouter is the better choice
什么时候 OpenRouter 更适合
OpenRouter is usually the better fit when speed matters more than infrastructure ownership, especially for prototypes and fast product iteration.
当上线速度比基础设施所有权更重要时，OpenRouter 通常更适合，尤其适合原型和快速产品迭代。
LiteLLM and OpenRouter help you decide how to reach models. They do not, by themselves, solve the next production-agent problem: how the agent discovers and calls reliable live data, verified tools, document processors, financial intelligence, and external services.
LiteLLM 和 OpenRouter 解决的是如何调用模型。它们本身并不解决生产级 Agent 的下一层问题：Agent 如何发现并调用可靠的实时数据、已验证工具、文档处理、金融情报和外部服务。
QVeris is the complementary capability routing layer. The stack is not LiteLLM versus QVeris or OpenRouter versus QVeris. It is model gateway plus QVeris.
QVeris 是互补的能力路由层。正确理解不是 LiteLLM vs QVeris，也不是 OpenRouter vs QVeris，而是模型网关 + QVeris。
| If you need...如果你需要…… | Use...选择 | Reason原因 |
|---|---|---|
| Self-hosted model gateway自托管模型网关 | LiteLLM | You control deployment, routing, keys, and provider configuration.您可以控制部署、路由、密钥和提供商配置。 |
| Fastest path to many models最快接入多种模型 | OpenRouter | The managed API removes gateway operations from your team.托管 API 让团队无需承担网关运维工作。 |
| Compliance and custom policies合规与自定义策略 | LiteLLM | Self-hosting and custom controls are easier to align with internal policy.自托管和自定义控制更容易满足内部政策要求。 |
| Low-friction experimentation低门槛实验 | OpenRouter | Developers can try many models without building infra first.开发者无需先搭建基础设施即可尝试多种模型。 |
| Agents that need live tools需要实时工具的 Agent | QVeris | Capability routing gives the model current data and real actions.能力路由为模型提供最新数据和真实执行能力。 |
Is LiteLLM better than OpenRouter?
LiteLLM 比 OpenRouter 更好吗？
LiteLLM is better for self-hosting and control. OpenRouter is better for managed convenience and fast access to many models.
LiteLLM 更适合自托管和控制权，OpenRouter 更适合托管便利和快速接入多模型。
Is OpenRouter self-hosted?
OpenRouter 可以自托管吗？
OpenRouter is a managed model gateway and aggregator. If self-hosting is a hard requirement, LiteLLM is usually the more natural fit.
OpenRouter 是托管式模型网关和聚合器。如果自托管是硬性要求，LiteLLM 通常更自然。
Can LiteLLM and OpenRouter work together?
LiteLLM 和 OpenRouter 能一起用吗？
Some teams combine gateways, but the simpler question is what each layer owns. Keep model routing clear, then add QVeris for capability calls.
有些团队会组合多个网关，但更重要的是明确每一层负责什么。模型路由保持清晰，再用 QVeris 补能力调用。
Do I need QVeris with LiteLLM or OpenRouter?
用了 LiteLLM 或 OpenRouter 还需要 QVeris 吗？
If your app only chats with a model, maybe not. If your agent needs live data, verified tools, or auditable execution, QVeris solves a different layer.
如果应用只是和模型聊天，也许不需要。如果 Agent 需要实时数据、已验证工具或可审计执行，QVeris 解决的是另一层问题。
