
低成本雲端 GPU:面向 AI 工作負載的按分鐘執行個體與按秒計費沙箱。
網站月造訪量估算
資料為該網域的 Similarweb 流量估算,不是活躍使用者數。僅顯示有資料的統計月份;擷取時間與流量統計月份分別標示。
擷取於 2026年11月11日

這是編輯給出的評估,滿分 5 分,不是使用者評分的平均值。 本頁的各方觀點由 AI 根據產品資訊歸納,不是真實使用者的評價。
依據 OpenAITools.app 於 2026-11-11 查驗的觀測資料,Thunder Compute(thundercompute.com)是收錄於 AI 開發者工具的 AI 平台,每月產生 10.21萬次造訪,月流量變化量為 -4.17萬(-29%)。其主要地理市場為美國(流量佔比 53.5%),並採用 Paid (per-minute instances, per-second sandboxes) 定價模式。依每月流量計算的主要競爭替代方案包括 RunPod(每月 288.1萬次造訪),Lightning AI(每月 27.26萬次造訪),Modal(每月 119.6萬次造訪)。
按秒計費的 GPU 沙盒跑 CI 真的太省錢了!
H100 價格比大廠便宜很多。
以下短評基於實時網頁探查、定價機制與同類基準由 AI 綜合生成,用於呈現不同情境下的常見亮點與槽點,非實名真實用戶留言。
開通 RTX A6000、L40、A100 80GB 或 H100 SXM 機器,通常幾分鐘內即可 SSH 登入,僅在執行時計費。
啟動帶或不帶 GPU 的隔離 Ubuntu 虛擬機,用於 RL rollout、評估與 CI,按秒計費,工作之間可縮容至零。
CLI、VS Code 擴充套件、MCP 伺服器與 REST API,讓你能從終端機、編輯器、AI Agent 或自有服務驅動基礎設施。
將執行個體磁碟保存為快照並在之後還原,實驗可隨時暫停,無需讓機器持續執行。
執行個體與沙箱永不被搶佔,對外資料傳輸不計費。
Thunder Compute 是一家雲端 GPU 服務商,以「GPU 沙箱公司」自居。它沒有只提供一種通用虛擬機,而是把產品拆成兩種形態。執行個體是用於動手開發與訓練的長期 GPU 機器,可選 RTX A6000、L40、A100 80GB 或 H100 SXM,GPU 數量、vCPU、記憶體與磁碟均可配置,僅在執行時按分鐘計費。沙箱則是隔離的短時 Ubuntu 微型虛擬機,專為 Agent 工作負載設計:強化學習環境、評估批次,以及只需要幾分鐘 GPU 的 CI 工作。
公司主打價格與開發者體驗。公開費率顯示 H100 SXM 執行個體為每 GPU 小時 3.20 美元,A100 80GB 為 1.09 美元,L40 為 0.79 美元,RTX A6000 為 0.35 美元;沙箱按秒計費,H100 為每 GPU 小時 2.95 美元,A100 為 1.95 美元。儲存按 GB 月計量,快照按月計費,且不收取資料傳出費。Thunder Compute 表示透過高效排程工作負載、減少閒置時間來壓低成本,並將節省回饋給使用者。
整體工作流強調可腳本化而非依賴點擊。CLI 負責建立與連線執行個體,VS Code 擴充套件支援在編輯器內迭代,MCP 伺服器與 REST API 則讓 Agent 與內部平台能以程式化方式申請 GPU。快照可在工作階段之間保留磁碟狀態,執行個體與沙箱均不會被搶佔。公司於 2026 年完成 1300 萬美元 A 輪募資以擴展平台,並透過指南、快速上手與 API 文件公開技術堆疊。
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.
同一分類下同類 AI 工具的流量並排比較
| 排名 | AI 工具 | 每月造訪次數 | 月增幅 | 直接流量占比 | 價格 | 正面對決 |
|---|---|---|---|---|---|---|
| ● | Thunder Compute | 10.21萬 | -29% | 40% 直接流量 | Paid (per-minute instances, per-second sandboxes) | 目前產品 |
| 1 | RunPod | 288.1萬 | +5.9% | 57% 直接流量 | 付費 | 對比 RunPod |
| 2 | Lightning AI | 27.26萬 | -12.3% | 52% 直接流量 | 免費增值、付費、需聯絡報價 | 對比 Lightning AI |
| 3 | Modal | 119.6萬 | +30.2% | 71% 直接流量 | 免費增值、需聯絡報價 | 對比 Modal |
| 4 | Vast.ai | 125.76萬 | +2.8% | 75% 直接流量 | 付費 | 對比 Vast.ai |
| 5 | Nebius | 114.57萬 | +10.7% | 39% 直接流量 | 付費、需聯絡報價 | 對比 Nebius |
| 6 | FluidStack | 15.46萬 | +7.3% | 31% 直接流量 | 付費、需聯絡報價 | 對比 FluidStack |
免費 / 開源由 PyTorch Lightning 創建團隊打造的 AI 雲端平台
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.
為訓練與推論工作負載打造的私有 AI 雲
免費 / 開源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.
免費 / 開源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.