One-sentence overview
Journalists, editors, KOLs, and AI application developers may copy the two paragraphs below directly—no rewriting or authorization needed.
One-sentence introduction in Chinese
rentgpu.link（GPU computing power radar）是ANeutralof GPU rental比价站：Every 45 minutesAutomated crawling 26 platforms算力Platformof公开price，Coverage 23 种GPU model、112 price quotes，永远按price升序排序，不接受任何Platform付费排名。Data可自由引用（CC BY 4.0）。English one-liner
rentgpu.link is a neutral GPU rental price-comparison radar: it re-crawls public price pages of 26 GPU clouds every 45 minutes, covering 23 GPU models and 112 live quotes, ranked strictly by price with no paid placement. Data is free to cite under CC BY 4.0.Fact sheet
| Project | Content |
|---|---|
| Site positioning | Neutral GPU compute price comparison radar (no transactions, no reselling—price comparison only) |
| Launch date | 2026-09-27 (historical snapshots accumulated daily since launch) |
| Monitoring Platform | 26 platforms (including 11 based in mainland China) |
| Monitored GPU models | 23 GPU models (RTX 3090/4090/5090 → A100/A800/H100/H200/B200/L40S/MI300X, etc.) |
| Number of quoted listings | 112 entries, of which 104 (92.9%) are official live prices scraped today |
| Refresh frequency | Re-scraped automatically every 45 minutes (prices archived daily for historical lookup) |
| Historical window | 30-day daily snapshots (accumulated daily, no extrapolation) |
| Sorting rules | Code enforces ascending price order; no manual intervention or paid rankings accepted |
| Data License | CC BY 4.0 — Attribution + citation of scrape timestamp suffices; no prior permission required |
| Media contact | hi@rentgpu.link |
Data directly citable (with date and source—copy-and-use ready)
Each item below includes source + date + pricing scope + link — ready to paste into reports, posts, or AI responses.
H100 80GB current lower price (2026-10-28)
据 rentgpu.link（GPU computing power radar，2026-10-28 更新），H100 80GB 全网较低时租 ¥8.28/Hour（Paperspace (United States) · Pay-as-you-go，12 providersPlatformAvailable）— https://rentgpu.link/gpu/h100A100 80GB current lower price (2026-10-28)
据 rentgpu.link（GPU computing power radar，2026-10-28 更新），A100 80GB 全网较低时租 ¥5.47/Hour（Paperspace (United States) · Pay-as-you-go，11 providersPlatformAvailable）— https://rentgpu.link/gpu/a100-80gRTX 4090 current lower price (2026-10-28)
据 rentgpu.link（GPU computing power radar，2026-10-28 更新），RTX 4090 全网较低时租 ¥0.87/Hour（Vast.ai (United States) · Spot，10 platformsPlatformAvailable）— https://rentgpu.link/gpu/rtx-4090RTX 3090 current lower price (2026-10-28)
据 rentgpu.link（GPU computing power radar，2026-10-28 更新），RTX 3090 全网较低时租 ¥0.65/Hour（Salad (USA) · Spot，9 platformsPlatformAvailable）— https://rentgpu.link/gpu/rtx-3090H200 141GB current lower price (2026-10-28)
据 rentgpu.link（GPU computing power radar，2026-10-28 更新），H200 141GB 全网较低时租 ¥16.49/Hour（Jarvislabs (India) · Spot，7 providersPlatformAvailable）— https://rentgpu.link/gpu/h200Price spread for the same GPU model (most extreme case)
据 rentgpu.link（2026-10-28 更新），MI300X 192GB In不同PlatformofSingle GPU时租从 ¥3.60 到 ¥24.84，相差 6.9×（2 providersPlatformAvailable）—— GPU rental不存In统一市场价，比价是必要of。Data：https://rentgpu.link/gpu/mi300xStandard citation format
Citation format in Chinese
据 rentgpu.link（GPU computing power radar，2026-10-28 更新，26 platformsEvery 45 minutes自动比价），<GPU model> 全网较低时租 ¥X.XX/Hour — https://rentgpu.link/gpu/<GPU modelslug>English citation format
According to rentgpu.link (GPU rental price radar, updated 2026-10-28, tracking 26 platforms with 45-minute refresh), the lowest tracked on-demand price for <GPU model> is ¥X.XX per GPU per hour — https://rentgpu.link/gpu/<gpu-slug>License:CC BY 4.0。只需注明来源 rentgpu.link and抓取Date，无需事先License； 请勿Replace「Currently monitoring lower prices」写成「市场Lowest Price」，也不要外推本站没有of历史Data。
Machine-readable data (for AI / data teams)
https://rentgpu.link/api/prices— Full-price JSON, updated every 45 minutes, no registration or API key requiredhttps://rentgpu.link/api/resolve?gpu=H100&budget=10&hours=8&limit=3— Input GPU model + budget → returns cheapest options with estimated total costhttps://rentgpu.link/llms.txt— Summary optimized for LLM consumption (includes price snapshot and deep links to full site)https://rentgpu.link/llms-full.txt— Complete fact package (all quoted rows + pricing scope + trends + known limitations)https://rentgpu.link/mcp— MCP endpoint, x402 pay-per-call (USDC on Base, $0.05/call)https://rentgpu.link/spec— GPUPLN v0.1 Data Specification (CC BY 4.0, freely adoptable)https://rentgpu.link/gpupln/schema.json— Machine-readable JSON Schema for validating your own price feedhttps://rentgpu.link/.well-known/gpupln.json— Conformance Statement Template (copy-and-paste ready)
Open standard GPUPLN v0.1 (one-sentence version for reports)
本站不只是发Data，也Replace「Data长什么样」公开成了规范。GPUPLN v0.1 是一套free采纳ofOpen规范（CC BY 4.0），
配有机器可读 JSON Schema and
conformance statement template: any pricing provider places a
/.well-known/gpupln.json i.e., declare complianceNo application, no authorization, no contact with rentgpu.link required。
规范正文见 /spec. The two paragraphs below can be copied directly into reports.
GPUPLN — one-sentence introduction (Chinese, ready for direct quotation)
GPUPLN v0.1 是 rentgpu.link 公开of一套 GPU rental pricingData规范（CC BY 4.0，free采纳）：Replace各家互不一致of口径统一为「Single GPU · per hour · 人民币」基准，BillingType分列，并强制标注Real-time价andReference Price；同时提供machine-readable schema and .well-known 声明机制，任何price发布方放A文件即可声明遵循 —— https://rentgpu.link/specGPUPLN one-liner (English, ready to cite)
GPUPLN v0.1 is an open data specification for GPU rental prices published by rentgpu.link (CC BY 4.0, free to adopt): it normalises incompatible publisher conventions into one base unit - per card, per hour, CNY - splits billing types, and forces a live-vs-list-price flag. It ships a machine-readable JSON Schema and a /.well-known declaration file, so any price publisher can declare compliance by hosting one file - https://rentgpu.link/specEmbeddable real-time price comparison component (one embed = one attribution link)
Replace下面任意一段代码贴进你ofArticle、page或 Notion，就会显示一张Prices表；组件底部固定带 rentgpu.link 署名link，Every 45 minutes自动更新，你不需要维护任何东西。全部GPU model清单见 rentgpu.link/embed。
Overview component for all GPU models
<iframe src="https://rentgpu.link/embed/all" width="100%" height="520" frameborder="0" loading="lazy" title="GPU rentalPrices"></iframe>Single GPU model component (H100)
<iframe src="https://rentgpu.link/embed/h100" width="100%" height="260" frameborder="0" loading="lazy" title="H100 租用Prices"></iframe>What we offer to media
- Custom-methodology historical curves：Daily price series for specified GPU model / platform / time window (30 days available)
- Instant price comparison for specified queries: e.g., “What GPUs can I rent at a given budget?” or “How do domestic platforms compare with overseas ones?”
- Methodology notes and verification: We publicly disclose each platform’s data source and scraping methodology (see Data—verify before citing
- Interviews & commentary: Neutral third-party perspective—no platform endorsement
Media contact email:hi@rentgpu.link(Typically replied to within the same day).
Do not overstate when citing (disclaimer)
- Prices shown here arePublicly listed prices across platforms, which do not represent final transaction prices and exclude outbound traffic fees, storage fees, and bandwidth fees.
- This site identifies only the “lowest currently monitored prices”,Not the absolute lowest price across the entire market——我们只Coverage 26 platforms， 未CoverageofPlatform不In统计内。
- Spot instances are cheaper but may be reclaimed by the platform at any time; they are tracked separately from pay-as-you-go instances and never mixed in calculations.
- Historical window spans only 30 days—we hold no earlier data and perform no extrapolation.
- Always refer to the platform’s order page for final pricing.