跳转到内容
收件箱
服务商
新闻
GPU
数据中心
可靠性
宕机了吗
变更
查询
成绩单
域名注册商
市场
搜索服务商、GPU、国家、数据中心…
⌘
K
CNY
中
收件箱
服务商
新闻
GPU
数据中心
可靠性
宕机了吗
变更
查询
成绩单
域名注册商
市场
推理API
在浏览器中即时筛选数千家服务商。
列表
地图
关系
比较
复制链接
链接已复制
已保存视图
1
热度
29 家服务商
推理API
服务商
类型
总部
地区
GPU
起价
GPU起价
Nebius
2004
Amsterdam-based AI cloud with GB300/B300/H200 capacity and Token Factory inference
GPU云
+1
NL
+2
GB300
B300
+6
—
¥10.36
/GPU时
immers.cloud
2018
Moscow GPU cloud with immersion-cooled servers and per-second billing
GPU云
+3
RU
5090
H200
+15
—
¥2.91
/GPU时
Denvr Dataworks
2007
Canadian AI cloud with sovereign GPU compute in Canada and the US
GPU云
+1
CA
GB200
B200
+8
—
¥9.37
/GPU时
Cloud.ru
2019
Sber's cloud provider (ex-SberCloud) with Evolution IaaS/PaaS and GPU rental
GPU云
+3
RU
B300
H100
+5
—
¥23.92
/GPU时
Intelion Cloud
2024
Moscow GPU cloud with per-second billing, H200/H100 SXM and 48GB RTX 4090 cards
GPU云
+2
RU
H200
H100
+6
—
¥3.99
/GPU时
Together AI
2022
Open-model inference API plus self-serve and reserved NVIDIA GPU clusters
GPU云
+1
US
GB300
B300
+4
—
¥26.70
/GPU时
fal
2021
Generative media platform with pay-per-use model APIs and serverless GPUs
无服务器
+1
US
B300
GB200
+5
—
¥26.77
/GPU时
FriendliAI
2021
Inference platform with model APIs, per-second dedicated GPUs and bring-your-own-GPU
无服务器
+1
B300
B200
+4
—
¥26.77
/GPU时
Cirrascale
2010
San Diego AI cloud with NVIDIA, AMD, Qualcomm and Tenstorrent systems on term pricing
GPU云
+1
US
B200
PRO 6000
+8
—
¥25.77
/GPU时
Hugging Face
2016
AI model hub selling GPU Spaces, dedicated Inference Endpoints and routed inference
无服务器
+1
US
B200
PRO 6000
+8
≈
¥161
/月
¥5.35
/GPU时
Hyperbolic
2021
Open-access AI cloud with on-demand H100/H200/B200 and reserved clusters
GPU云
+2
B200
H200
+1
—
¥8.63
/GPU时
DeepInfra
2013
Low-cost model inference API plus dedicated A100-to-B300 GPUs billed by the minute
无服务器
+1
US
B300
B200
+4
—
¥5.96
/GPU时
Fireworks AI
2022
Fast inference and fine-tuning platform with per-second on-demand GPU deployments
无服务器
+1
US
GB300
B300
+3
—
¥53.54
/GPU时
Baseten
2019
Inference platform for dedicated model deployments and model APIs, billed per minute
无服务器
+1
US
B200
H100
+5
—
¥5.68
/GPU时
GMI Cloud
2024
NVIDIA Cloud Partner renting H100/H200/B200 and GB200 from its own data centers
GPU云
+1
US
GB300
GB200
+3
—
¥17.40
/GPU时
Groq
2016
Inference neocloud built on its own LPU accelerators, serving models via GroqCloud
推理
US
+5
—
—
NetMind
2023
Model APIs plus on-demand H100/A100/RTX 4090 GPUs from a distributed fleet
推理
+1
GB
H100
4090
+2
—
—
Novita AI
2023
Model APIs plus on-demand, spot and serverless GPU instances and bare metal
GPU云
+2
H200
H100
—
—
SiliconFlow
2003
Chinese AI inference platform serving open models with fine-tuning and reserved GPUs
推理
CN
H200
H100
+1
—
—
Bitdeer AI
2024
Singapore AI cloud from Bitdeer with GB300/GB200 NVL72 and serverless models
GPU云
+1
SG
GB300
GB200
+3
—
—
Parasail
2018
AI deployment network for serverless, dedicated and batch inference across GPU suppliers
推理
B300
B200
+4
—
—
Theta EdgeCloud
2024
Theta Labs' hybrid cloud-edge GPU platform for compute, inference and AI agents
GPU云
+2
US
5090
H200
+10
—
—
Nscale
2024
Vertically integrated AI cloud with hydro-powered data centers in Norway and the UK
GPU云
+1
GB
+1
—
—
Cerebras
2021
Wafer-scale chipmaker running one of the fastest LLM inference clouds
推理
US
—
—
Chutes
2024
Open-source, decentralized serverless platform for running open models at scale
无服务器
+2
—
—
SambaNova
2017
Dataflow-chip company serving fast open-model inference through SambaCloud
推理
US
—
—
Replicate
2019
Run and deploy open models through an API, billed per second; now part of Cloudflare
无服务器
+1
US
H200
H100
+4
—
¥5.42
/GPU时
Lepton AI
2023
AI cloud platform acquired by NVIDIA and relaunched as DGX Cloud Lepton
无服务器
+1
GB300
GB200
+1
—
—
OctoAI
2016
Model-serving platform (formerly OctoML) acquired by NVIDIA and shut down in 2024
推理
5080
L4
—
—
正在加载索引…