
低成本云 GPU:面向 AI 工作负载的按分钟实例与按秒计费沙箱。
网站月访问量估算
数据为该域名的 Similarweb 流量估算,不是活跃用户数。仅展示有数据的统计月份;采集时间与流量统计月份分别标注。
采集于 2026年11月11日

这是编辑给出的评估,满分 5 分,不是用户打分的平均值。 本页的各方观点由 AI 根据产品信息归纳,不是真实用户的评价。
根据 OpenAITools.app 于 2026-11-11 核对的流量数据,Thunder Compute(thundercompute.com)是一个收录在 AI Developer Tools 下的 AI 平台,月访问量为 10.21万,月度环比流量变化为 -4.17万(-29%)。其主要地理市场为美国(53.5% 流量份额),采用 Paid (per-minute instances, per-second sandboxes) 定价模式。按月流量计算的主要竞争替代方案包括 RunPod(288.1万次访问/月),Lightning AI(27.26万次访问/月),Modal(119.6万次访问/月)。
按秒计费的GPU沙盒跑自动化测试爽到飞起。
价格比AWS和GCP香太多了。
偶尔会出现抢不到卡的情况。
SSH连上去直接开干,很纯粹的开发者工具。
以下短评基于实时网页探查、定价机制与同类产品基准由 AI 综合模拟生成,旨在呈现不同使用场景下的常见槽点与亮点,非实名真实用户评价。
开通 RTX A6000、L40、A100 80GB 或 H100 SXM 机器,通常几分钟内即可 SSH 登录,只在运行时计费。
启动带或不带 GPU 的隔离 Ubuntu 虚拟机,用于 RL rollout、评估与 CI,按秒计费,任务之间可缩容至零。
CLI、VS Code 扩展、MCP 服务器与 REST API,让你能从终端、编辑器、AI Agent 或自有服务驱动基础设施。
将实例磁盘保存为快照并在之后恢复,实验可随时暂停,无需让机器持续运行。
实例与沙箱永不被抢占,出站数据传输不计费。
Thunder Compute 是一家云 GPU 服务商,以「GPU 沙箱公司」自居。它没有只提供一种通用虚拟机,而是把产品拆成两种形态。实例是用于动手开发与训练的长期 GPU 机器,可选 RTX A6000、L40、A100 80GB 或 H100 SXM,GPU 数量、vCPU、内存与磁盘均可配置,仅在运行时按分钟计费。沙箱则是隔离的短时 Ubuntu 微虚拟机,专为 Agent 工作负载设计:强化学习环境、评估批次,以及只需要几分钟 GPU 的 CI 任务,用完即弃。
公司主打价格与开发者体验。公开费率显示 H100 SXM 实例为每 GPU 小时 3.20 美元,A100 80GB 为 1.09 美元,L40 为 0.79 美元,RTX A6000 为 0.35 美元;沙箱按秒计费,H100 为每 GPU 小时 2.95 美元,A100 为 1.95 美元。存储按 GB 月计量,快照按月计费,且不收取数据出网费。Thunder Compute 表示通过高效调度工作负载、减少空闲时间来压低成本,并将节省转嫁给用户。
整个工作流强调可脚本化而非依赖点击。CLI 负责创建与连接实例,VS Code 扩展支持在编辑器内迭代,MCP 服务器与 REST API 则让 Agent 和内部平台能够以编程方式申请 GPU。快照可在会话之间保留磁盘状态,实例与沙箱均不会被抢占。公司于 2026 年完成 1300 万美元 A 轮融资以扩展平台,并通过指南、快速上手与 API 文档公开技术栈。
Rent a GPU by the minute, attach VS Code, and iterate on CUDA code or fine-tuning without long-term commitments.
Run thousands of short rollouts and evaluations in isolated GPU VMs billed by the second, scaling to zero between batches.
Provision GPUs from inside a coding agent or CI system via the MCP server, CLI, or REST API instead of a web console.
Use a single snapshot-backed A100 overnight for training jobs at a fraction of the cost of major cloud list prices.
同一分类下同类 AI 工具的流量并排对比
| 排名 | AI 工具 | 月访问量 | 月度环比增长 | 直接流量占比 | 定价 | 头对头 |
|---|---|---|---|---|---|---|
| ● | Thunder Compute | 10.21万 | -29% | 40% 直接流量 | Paid (per-minute instances, per-second sandboxes) | 当前产品 |
| 1 | RunPod | 288.1万 | +5.9% | 57% 直接流量 | 付费 | 对比 RunPod |
| 2 | Lightning AI | 27.26万 | -12.3% | 52% 直接流量 | 免费增值,付费,定价请咨询 | 对比 Lightning AI |
| 3 | Modal | 119.6万 | +30.2% | 71% 直接流量 | 免费增值,定价请咨询 | 对比 Modal |
| 4 | Vast.ai | 125.76万 | +2.8% | 75% 直接流量 | 付费 | 对比 Vast.ai |
| 5 | Nebius | 114.57万 | +10.7% | 39% 直接流量 | 付费,定价请咨询 | 对比 Nebius |
| 6 | FluidStack | 15.46万 | +7.3% | 31% 直接流量 | 付费,定价请咨询 | 对比 FluidStack |
免费 / 开源由 PyTorch Lightning 创建团队打造的 AI 云平台
免费 / 开源面向 AI 与数据负载的无服务器云基础设施
从全球市场按秒租用云 GPU
Nebius is an AI-focused cloud platform that gives teams non-virtualized NVIDIA GPU clusters, managed MLOps tooling, and a serverless inference layer so they can move from experiment to production on one stack.
Nebius is an AI-focused cloud platform that gives teams non-virtualized NVIDIA GPU clusters, managed MLOps tooling, and a serverless inference layer so they can move from experiment to production on one stack.

FluidStack builds and operates large-scale compute infrastructure for AI, including gigawatt-class data centers and GPU clusters used for training and inference. The company designs its own power, cooling and data center systems, onshores manufacturing and deploys solar generation to supply the power its clusters need. Rather than a self-serve cloud console, FluidStack works with AI labs and enterprises to provision large reserved clusters, with orchestration tooling for the ML frameworks and schedulers those teams already use.
FluidStack builds and operates large-scale compute infrastructure for AI, including gigawatt-class data centers and GPU clusters used for training and inference. The company designs its own power, cooling and data center systems, onshores manufacturing and deploys solar generation to supply the power its clusters need. Rather than a self-serve cloud console, FluidStack works with AI labs and enterprises to provision large reserved clusters, with orchestration tooling for the ML frameworks and schedulers those teams already use.

Massed Compute is a cloud GPU provider offering on-demand NVIDIA instances, single-tenant bare metal servers, large InfiniBand node clusters, and a programmatic REST API and MCP server. It targets AI training, inference, machine learning, data analytics, and VFX rendering workloads.
Massed Compute is a cloud GPU provider offering on-demand NVIDIA instances, single-tenant bare metal servers, large InfiniBand node clusters, and a programmatic REST API and MCP server. It targets AI training, inference, machine learning, data analytics, and VFX rendering workloads.
面向训练与推理工作负载的私有 AI 云
免费 / 开源编排全球闲置消费级 GPU 的分布式 GPU 云
免费 / 开源General Compute is an inference cloud built around alternative AI silicon rather than standard GPU fleets. The company buys racks of accelerators, racks them in colocation data centres, and handles the cooling, power, and networking itself. On top of that hardware it runs a disaggregated inference stack that splits each request: prompt processing runs on NVIDIA GPUs for prefill, while token generation runs on faster decode chips such as Cerebras, SambaNova, Positron AI, and d-Matrix. Customers get bare metal access or a tokens-per-second SLA.
免费 / 开源General Compute is an inference cloud built around alternative AI silicon rather than standard GPU fleets. The company buys racks of accelerators, racks them in colocation data centres, and handles the cooling, power, and networking itself. On top of that hardware it runs a disaggregated inference stack that splits each request: prompt processing runs on NVIDIA GPUs for prefill, while token generation runs on faster decode chips such as Cerebras, SambaNova, Positron AI, and d-Matrix. Customers get bare metal access or a tokens-per-second SLA.