AI gateway status checks
Manually verified status, latency references, and /models support for each gateway. This is not real-time monitoring — still load-test and configure backup providers before production.
- Small amounts
- Trial first
Status Table
Availability and latency reference
Latency is a reference value measured at verification time; region, network, model, and account quota all change the real result.
| Gateway | Latency reference | /models | Checked on | Notes |
|---|---|---|---|---|
| CodexHot | Not verified yet | Public group guide; check the dashboard for live models | 2026-10-07 | Public integration, SDK, and group guides were accessible. Model tests are platform-reported; AI Beyond has not tested paid inference or latency. The external voucher shop returned 403, so payment methods remain unverified. No dedicated public status page has been verified. |
| Infer by Flow7 | Not verified yet | /api/public/catalog | 2026-10-01 | The public catalog, status, docs, and OpenAPI contract were accessible; the status API reported operational. AI Beyond has not tested paid inference or latency. Check model and policy availability immediately before use. |
| OpenRouter | 1323 ms | /models supported | 2026-09-15 | Docs, model catalog, and pricing page are fairly complete; a good candidate for multi-model routing. |
| Vercel AI Gateway | 1943 ms | Platform model catalog | 2026-09-15 | Suits the Vercel AI SDK and frontend/full-stack projects; an engineering gateway rather than a prepaid relay. |
| 302.AI | 2559 ms | See the console | 2026-09-15 | Broad coverage, but pricing and model availability change quickly — test with a small balance. |
| OhMyGPT | 1757 ms | See the console | 2026-09-15 | A China-facing prepaid service, fine for low-value testing; avoid parking a large balance long term. |
| SiliconFlow | 334 ms | Model hub supported | 2026-09-15 | A Chinese model platform, good for DeepSeek, Qwen, embeddings, speech, and image capabilities. |
| AI/ML API | 1406 ms | Model list in the docs | 2026-09-15 | An overseas multi-model aggregator, suited to developers who can use overseas billing. |
| Portkey AI Gateway | 1923 ms | Provider routing | 2026-09-15 | Leans toward enterprise governance, logging, and routing; good for managing multiple providers. |
| LiteLLM Proxy | Depends on your deployment | Self-hosted configuration | 2026-09-15 | An open-source proxy layer; real latency and availability depend on your deployment and upstream providers. |
| One API | Depends on your deployment | Self-hosted configuration | 2026-09-15 | An open-source channel-management project, suited to internal team or webmaster self-hosting. |
| New API | Not verified yet | See the docs | Not verified yet | Public operation means owning key security, channel compliance, billing risk-control, and user-data protection. |
| Hezu Bus | Not verified yet | See the docs | Not verified yet | Public docs are limited; before production, verify model sources, billing rules, log retention, and refund terms. |
| Sublyx | Not verified yet | See the docs | Not verified yet | For services involving top-up and accounts, check the terms carefully; avoid uploading sensitive data or keeping large balances. |
| Msairelay | Not verified yet | See the docs | Not verified yet | Public info is limited; first verify model names, error codes, log policy, and whether failed requests are billed. |
| OpenAI Official API | Not verified yet | See the docs | Not verified yet | China networking, account region, and payment must be solved yourself; no multi-provider aggregation. |
| Claude API (Anthropic) | Not verified yet | See the docs | Not verified yet | Not an OpenAI-compatible relay; payment, region, and network availability need separate evaluation. |
| Google AI Studio | Not verified yet | See the docs | Not verified yet | Google's native interface differs from the OpenAI-compatible format; test China access and payment terms yourself. |
| xAI Grok API | Not verified yet | See the docs | Not verified yet | Models, region, and API policy change fast; recheck rate limits, price, and data terms before launch. |
| Groq | Not verified yet | See the docs | Not verified yet | Model coverage leans toward the open ecosystem; verify China networking, rate limits, and enterprise support separately. |
| Together AI | Not verified yet | See the docs | Not verified yet | Overseas billing and network availability need self-testing; closed-model coverage is less than full aggregators. |
| Fireworks AI | Not verified yet | See the docs | Not verified yet | More an overseas open-model infrastructure; assess China direct-connection experience and enterprise billing. |
| DeepInfra | Not verified yet | See the docs | Not verified yet | Limited closed-model coverage; before production verify rate limits, availability zones, and billing of failed requests. |
| Xigua API | Not verified yet | See the docs | Not verified yet | Public info is scarce; focus on verifying prices, model sources, log policy, and balance-refund rules. |
| Wuai API | Not verified yet | See the docs | Not verified yet | Public docs are limited; not recommended for production or sensitive data before verification. |
| Apito (formerly ClaudeAPI) | Not verified yet | See the docs | Not verified yet | Third-party Claude forwarding requires verifying data retention, channel details, refunds, and model availability. |
| laozhang.ai | Not verified yet | See the docs | Not verified yet | Prices, model names, and availability may change; keep official or backup providers for production. |
| AiHubMix | Not verified yet | See the docs | Not verified yet | Available models and prices on multi-model platforms change fast; test important workflows item by item. |
| API2D | Not verified yet | See the docs | Not verified yet | Model coverage is concentrated; if you need Claude, Gemini, or domestic models, pair with another platform. |
| CloseAI | Not verified yet | See the docs | Not verified yet | Public docs and entity info need further verification; not recommended for sensitive or production data directly. |
| V3 API | Not verified yet | See the docs | Not verified yet | Public info is limited; before launch confirm model names, multipliers, logs, and failed-request billing. |
| GPTGod | Not verified yet | See the docs | Not verified yet | Low-price platforms especially require verifying stability, model sources, balance rules, and support responsiveness. |
| 0011.ai | Not verified yet | See the docs | Not verified yet | Public info is limited; verify actually supported models, tool-calling compatibility, and billing rules. |
| CoderPlan | Not verified yet | See the docs | Not verified yet | Price discounts, node latency, and tool compatibility should be verified with real requests. |
| HCAI | Not verified yet | See the docs | Not verified yet | Public docs are limited; test small amounts first and verify logs, refunds, and data-retention rules. |
| DeepSeek Official API | Not verified yet | See the docs | Not verified yet | Covers only DeepSeek's own models; peak-time availability and rate limits need separate evaluation. |
| Zhipu AI BigModel | Not verified yet | See the docs | Not verified yet | Not an overseas-model relay; if you target Claude/Gemini/OpenAI, pair with another provider. |
| Alibaba Cloud Model Studio | Not verified yet | See the docs | Not verified yet | Cloud-vendor interfaces, auth, and model names may differ from the overseas OpenAI ecosystem. |
| Volcengine Ark (Doubao) | Not verified yet | See the docs | Not verified yet | Mainly covers Doubao and Volcengine-ecosystem models; cross-provider routing needs separate design. |
| Kimi Open Platform (Moonshot AI) | Not verified yet | See the docs | Not verified yet | Mainly covers Moonshot/Kimi models; not suitable as a multi-provider aggregation platform. |
| MiniMax API | Not verified yet | See the docs | Not verified yet | Interfaces and model system differ from the OpenAI-compatible ecosystem; integrate via the official SDK/docs. |
| Baidu Qianfan | Not verified yet | See the docs | Not verified yet | The interface system may differ from the OpenAI-compatible ecosystem; verify SDK, auth, and model params before migrating. |
Method
Checks worth running yourself
- Call `/v1/models` or the platform model list to confirm your target model name exists.
- Test plain responses, streaming, function calling, embeddings, and image or speech endpoints separately.
- Record timeouts, 429s, insufficient balance, and retired-model errors — and confirm whether they are billed.
- Configure backup providers and fallback models so one gateway outage does not take production down.
Risk
Risk notice
This site only organizes information about AI API relay services; it makes no guarantee about third-party platforms' stability, price, data security, or availability. Judge the risk yourself before use, and avoid uploading sensitive data, personal privacy, trade secrets, or core API keys to unknown platforms.