FastGPU bronze
fastgpu.co
“Live GPU cloud price comparison across the whole market, as a REST API and MCP server. Compare live cloud GPU prices across RunPod, Vast.ai, Lambda, the hyperscalers and more. Describe your AI workload and see the GPU clouds that can run it.
a2a https://fastgpu.co talk to it https://fastgpu.co/.well-known/agent-card.json its cardwe checked this the operator says this
Verified by agenttru.st
Everything here is a check agenttru.st performed itself. Assurance, protocol, hosting and freshness are in the card above and are not repeated.
- Certificate
-
Issued by Let's Encrypt
domain-validated
Valid until 30 Dec 2026 as of our check on 16 Nov 2026.Control of the hostname was checked; nothing about who operates it.Key RSA 2048, signed SHA256-RSA, issued via YR1 under ISRG Root X1 β classical, as every publicly trusted certificate is today.
Names on this certificate (1)
fastgpu.co. - Connection
- TLS 1.3 Β· X25519MLKEM768 β post-quantum hybrid key exchange, agreed with our checker on 16 Nov 2026. What this means
- DANE / TLSA
- Not verified (TLSA query returned RCodeNameError)
- Discovery
- Well-known document
- AI-facing documents
-
Publishes
/llms.txtβ “FastGPU” a curated map of the site's content for language modelsFetched from this host during verification. Neither document is how this agent was discovered. - First seen
- 4 Oct 2026
View verification details
- Assurance
- bronze Bronze β agent card fetched over HTTPS with a valid certificate
- Protocols
- A2A verified by handshake or card fetch, not merely advertised
- Hosted in
- πΊπΈ US Β· Amazon.com, Inc. (AS16509)
- Last checked
- 22h ago
What this agent says it can do
Declared in the agent's own card. agenttru.st has not tested whether it completes any of these tasks β the operator of fastgpu.co controls every word below.
List cheapest GPU prices
One entry per GPU model with its cheapest live on-demand cloud rental price in USD per GPU-hour, the provider offering it, and how many providers rent it, from FastGPU's live price feed across marketplaces, neoclouds and hyperscalers (RunPod, Vast.ai, Lambda, AWS and more). Use it to compare GPU rental prices or to find where a named GPU is cheapest to rent. Each price is the provider's own per-GPU rate: cheapest_min_gpu_count says when it is only sold as a multi-GPU instance, and cheapest_billed_separately names what the provider bills on top of it (CPU, memory, storage), null when the rate includes CPU and memory. page_url links to the GPU's live price page with every offer. Rental prices only: it does not rent, reserve or launch GPUs. No key required.
Match a workload to the cheapest GPU
Ranks up to eight GPU configurations that can run a job at the lowest live price. Describe the job in query (an open-weight model to serve or fine-tune, a model size, or a GPU need) or with the structured fields. Each match gives the GPU count, the USD per hour for the whole configuration and a one-line reason; the answer also states the VRAM the job needs and how much cheaper the top match is than a hyperscaler running the same job. Use it to decide where to run a workload; use list_gpu_prices for the price of a named GPU. Open-weight models only: an API-only model such as GPT-4 returns no matches and a note. Ranked on price and fit alone: no provider pays for placement (each match still reports partner, a provider that pays FastGPU a referral). Hard limits (budget_usd_hr, max_total_usd with duration_hours, max_quote_age_minutes) drop every configuration that breaks them; limits reports what was dropped. Read-only: it does not rent, reserve or launch GPUs. No key required.
List daily GPU price fixings
The frozen daily record of GPU rental prices: one row per GPU model per UTC day, taken at 07:00 UTC from the live price snapshot and never revised. Each row carries the floor (the cheapest posted on-demand price outside the hyperscalers) and the provider that set it, the on-demand median across providers, the spot floor, the hyperscaler median, the provider count behind each figure, and the in-stock share as a numerator and a denominator. Prices are posted list prices in USD per GPU-hour, not transaction prices. Newest day first; filter by date or GPU and page with limit and offset. The method is published at /methodology/fixings. No key required.
List the GPU Availability Index
How often each GPU could actually be rented: one frozen row per GPU model per completed UTC day, computed from FastGPU's recorded stock history (each provider's own stock tier or rentable host count, on-demand and community offers only) and never revised. Each row gives the providers listing the GPU, the providers that exposed a stock signal, the providers that had it rentable at least once that day, the signalled and rentable stock checks, the rentable share (rentable / signalled; offers without a signal never count as in stock or sold out) and the marketplace host depth. Newest day first. The method is published at /gpu-availability#method. No key required.
Get GPU price volatility and spread
Price-risk statistics per GPU from FastGPU's own record of posted rental prices: realized volatility of the recorded daily floor (standard deviation of daily log changes, per day and annualized with the square root of 365, with the number of changes behind it) over the last 30 days and the full history, the largest one-day move, drawdown and run-up, the spread across providers right now (P10, median and P90 of each non-hyperscaler provider's lowest buyable price), which provider set the daily fixing's floor and how persistently, the basis between the on-demand median and the floor and between the hyperscaler median and the on-demand median, and confirmed list-price changes over 60 days. Posted prices, not transactions, and not a settlement reference. No key required.
List AWS Capacity Block offers
What AWS answers, every 6 hours, to "when can I start N GPUs and for how much": the latest run of read-only EC2 DescribeCapacityBlockOfferings calls across its GPU instance types (H100, H200, B200, B300, A100), default-enabled Regions and four shapes (1 instance for 1 day or 1 week, 8 instances for 2 or 13 weeks). Each row gives how many offers AWS returned, the earliest start with its upfront fee, and the lowest price per GPU-hour with its start; status 'error' rows carry AWS's error code. AWS prices Capacity Blocks by supply and demand and shows them only to signed-in accounts. Fees are USD for Linux. The response carries the method. No key required.
Estimate what a fine-tune costs on every GPU
For a LoRA, QLoRA or full fine-tune of an open language model on a dataset: the GPU memory it needs, and for every GPU that can hold it, how many of that GPU, how long the run takes and what the whole job costs at that GPU's cheapest rentable live price, billed under its provider's meter, cheapest job first. Time = 8 (full) or 6 (LoRA) x active parameters x tokens / (GPUs x the card's dense 16-bit peak x mfu); mfu is an assumption the caller picks. Each row carries page_url (the GPU's live prices) and rent_url (opens the provider through a FastGPU redirect that counts the click; partner rows carry FastGPU's referral code at the same price). Read-only. No key required.
List GPU price changes
FastGPU's price-change log: every list-price change its own record of posted prices confirms (a move between two prices that each held for two complete UTC days, at least a cent apart, on-demand and reserved, for providers that publish their own per-GPU rate) plus changes providers announce on their own pages, newest first. Each entry gives the provider, GPU, old and new price per GPU-hour, the change in percent, the day (or window) of the change and, for an announced change, the provider's page and wording. No key required.
Get every live offer (full feed)
Every normalized live offer (provider, GPU, offer type, region, per-GPU price, availability), for building tools or dashboards on the raw feed. Requires an API key. Filter by gpu, provider, type, limit.
Launch a GPU (route + provision)
Route a described job to the best in-stock connected provider (RunPod, Vast.ai, Hyperstack) and PROVISION a real GPU, billed by the second to your prepaid FastGPU wallet. Returns the instance; poll GET /api/v1/router/instances/{id} for the connection URL once it is running. Requires an API key with wallet balance.
List your GPU instances + balance
List your router GPU instances (status, connection, billed rate) and your wallet balance. Refreshes any live instance against its provider first, so status + connection resolve. Requires an API key.
Get one GPU instance
Get one instance's live status + connection URL. Poll this after launching until status is "running". Requires an API key.
Terminate a GPU instance
Terminate one instance: destroys the GPU on the provider and settles your wallet to the actual elapsed cost. Requires an API key.
Technical agent card
Copied from the agent's card. The operator controls these values; agenttru.st has not verified them.
- Provider
- FastGPU β what this agent says about itself; other agents claiming the same provider are not thereby related
- Protocol
- a2a
- Version
- 1.0.0
- Auth schemes
- apiKey
- Card completeness
-
a2a.proto v1.0 requires eight top-level fields. This card omits:
Missing fields do not affect listing β they describe how much the operator has published, not whether the agent was verified.
View all card details
- Capabilities
- pushNotifications stateTransitionHistory streaming
- Agent card
- https://fastgpu.co/.well-known/agent-card.json
Operate this agent and would rather not be listed? Request removal.