// Models · commercial services
Compliant commercial services
Where to buy model APIs, hosting, GPUs and data through channels with a clear authorisation basis. Not endorsements — check terms and regional availability yourself.
56 services7 groupschecked 2026-10-22Model API prices →
Compliant aggregators & routersAPI aggregation & routing
| Service | What it offers | Chinese |
|---|---|---|
| OpenRouteropenrouter.ai | Unified OpenAI-compatible API routing to 400+ models, pay as you go | — |
| Requestyrequesty.ai | Hosted LLM routing gateway, 600+ models | — |
| Portkeyportkey.ai | Hosted AI gateway with observability and guardrails; gateway is open source (MIT) | — |
| LiteLLMlitellm.ai | Open-source self-hosted LLM proxy (MIT core) plus enterprise edition | — |
| Vercel AI Gatewayvercel.com | Model gateway built into the Vercel platform | — |
| Cloudflare AI Gatewaydevelopers.cloudflare.com | Edge AI gateway (caching, rate limiting, logs) | — |
| Eden AIedenai.co | One API for many AI services (LLM, OCR, speech) | — |
| LLM Gatewayllmgateway.io | Open-source (AGPLv3) and hosted gateway, 200+ models | — |
| Unifyunify.ai | Multi-model routing | — |
| Respanrespan.ai | Gateway + observability + evaluation, 1,000+ models | — |
| Kong AI Gatewaykonghq.com | AI plugins for the Kong enterprise API gateway | — |
| Heliconehelicone.ai | AI gateway + LLM observability (open source) | — |
Cloud model hosting (MaaS)API aggregation & routing
| Service | What it offers | Chinese |
|---|---|---|
| Amazon Bedrockaws.amazon.com | Hosted Claude, Llama, Mistral and other models | — |
| Google Cloud Vertex AI (now Gemini Enterprise Agent Platform)cloud.google.com | Hosted Gemini, Claude and other models | — |
| Microsoft Foundry (formerly Azure AI Foundry)ai.azure.com | Hosted OpenAI, Claude and other models | — |
| Alibaba Cloud Model Studio (Bailian)bailian.aliyun.com | API for Qwen and third-party Chinese models; Token Plan | ✓ |
| Volcengine Arkvolcengine.com | API for Doubao and third-party models; Coding Plan | ✓ |
| Baidu AI Cloud Qianfancloud.baidu.com | API for ERNIE and third-party models | ✓ |
| Tencent Cloud Hunyuancloud.tencent.com | Hunyuan model API | ✓ |
| Qiniu AI model platformqiniu.com | Unified access to Chinese models, OpenAI/Anthropic-compatible | ✓ |
Open-source gateway softwareAPI aggregation & routing
GPU rentalOther commercial services
| Service | What it offers | Chinese |
|---|---|---|
| RunPodrunpod.io | GPU cloud (Pods / Serverless) | — |
| Lambdalambda.ai | GPU cloud and clusters | — |
| Vast.aivast.ai | GPU rental marketplace | — |
| CoreWeavecoreweave.com | Large-scale GPU cloud | — |
| Modalmodal.com | Serverless GPU compute | — |
| Hyperbolichyperbolic.ai | GPU rental + inference | — |
| AutoDLautodl.com | GPU rental in mainland China | ✓ |
| Gpusharegpushare.com | GPU rental in mainland China | ✓ |
| Matpoolmatpool.com | GPU rental in mainland China | ✓ |
| Featurizefeaturize.cn | Online GPU lab in mainland China | ✓ |
| GpuGeekgpugeek.com | GPU cloud, including overseas regions | ✓ |
Inference hostingOther commercial services
| Service | What it offers | Chinese |
|---|---|---|
| Together AItogether.ai | Open-model inference, fine-tuning and GPU clusters | — |
| Fireworks AIfireworks.ai | Inference + training platform | — |
| Replicate (acquired by Cloudflare)replicate.com | Hosted model runs via API (acquired by Cloudflare) | — |
| Basetenbaseten.co | Inference deployment platform | — |
| Groqgroq.com | Low-latency inference | — |
| Cerebrascerebras.ai | High-throughput inference | — |
| DeepInfradeepinfra.com | Serverless open-model inference | — |
| Novita AInovita.ai | Model API + GPU cloud | — |
| Hugging Face Inference Providershuggingface.co | One entry point to many inference providers | — |
| SiliconFlowsiliconflow.cn | API for Chinese and open models, dedicated instances, private deployment | ✓ |
| PPIOppio.com | Model API, GPU / Serverless, private deployment | ✓ |
| Infinigence AIinfinigence-ai.com | Heterogeneous compute + MaaS “token factory” | ✓ |
Fine-tuning & training servicesOther commercial services
| Service | What it offers | Chinese |
|---|---|---|
| Tinker (Thinking Machines)thinkingmachines.ai | Hosted LoRA training API | — |
| OpenPipe (now part of CoreWeave/W&B)openpipe.ai | Formerly hosted fine-tuning/RL; the old platform stopped new training and inference from 2026-07-30 | — |
| Predibase (site redirects to Rubrik)predibase.com | Formerly a LoRA fine-tuning platform; as of 2026-10-22 the site still redirects to rubrik.com, standalone service status unverified | — |
Data labelling & expert dataOther commercial services
| Service | What it offers | Chinese |
|---|---|---|
| Scale AIscale.com | Data labelling, evaluation, government and enterprise AI | — |
| Surge AIsurgehq.ai | RLHF and expert data | — |
| Mercormercor.com | Expert human data platform | — |
| Labelboxlabelbox.com | Labelling platform and data services | — |
| SpeechOceanspeechocean.com | Training data services (listed in Shanghai) | ✓ |
| Datatangdatatang.com | Training data services | ✓ |
| Testin Datatestin.cn | AI data collection and labelling | ✓ |
| Molar Datamolardata.com | AI data collection and labelling | ✓ |
Pricing notes and the authorisation basis for each service are recorded in Chinese on the Chinese version of this page.