SpotGPUs vs Vast.ai: −48% on the same GPUs
Across the 39 GPU models both catalogues list, our on-demand rate is on average 48% below the price published by Vast.ai, and our spot rate is -27% below theirs. Every figure below is dated and sourced.
- Models compared
- 39Present in both catalogues
- Average gap
- −48%Our on-demand against their published rate
- Biggest gap
- −75%H200 NVL — $0.99 against $4.03
- Recorded
- 15 Sep 2026From getdeploying.com and 4 other sources
The 39 GPUs you can rent from either of us.
Their column is the lowest published price we recorded for that GPU on 15 September 2026. Ours is today's catalogue rate, the one the console bills.
| GPU | Vast.aion-demand | SpotGPUson-demand | Gap | SpotGPUsspot | Source |
|---|---|---|---|---|---|
| B200 SXM180 GB HBM3e | $5.63spot $5.31 | $2.49 | −56% | $1.6932% off ours | getdeploying.com |
| H200 SXM141 GB HBM3e | $3.60spot $1.32 | $1.09 | −70% | $0.7531% off ours | getdeploying.com |
| H200 NVL141 GB HBM3e | $4.03spot $1.32 | $0.99 | −75% | $0.6930% off ours | cloud-gpus.com |
| H100 SXM80 GB HBM3 | $1.73spot $0.69 | $0.89 | −49% | $0.5538% off ours | gpucloudcost.com |
| H100 NVL94 GB HBM3 | $2.78spot $0.02 | $0.79 | −72% | $0.4938% off ours | computeprices.com |
| H100 PCIe80 GB HBM2e | $2.13spot $1.33 | $0.69 | −68% | $0.4535% off ours | console.vast.ai |
| RTX PRO 6000 Blackwell96 GB GDDR7 | $0.87spot $0.34 | $0.39 | −55% | $0.1951% off ours | getdeploying.com |
| A100 SXM 80GB80 GB HBM2e | $0.53spot $0.29 | $0.29 | −46% | $0.1934% off ours | gpucloudcost.com |
| RTX PRO 5000 Blackwell48 GB GDDR7 | $0.66spot $0.14 | $0.29 | −56% | $0.1548% off ours | getdeploying.com |
| L40S48 GB GDDR6 | $0.73spot $0.39 | $0.25 | −66% | $0.1540% off ours | gpucloudcost.com |
| A100 PCIe 80GB80 GB HBM2e | $0.54 | $0.19 | −64% | $0.0953% off ours | computeprices.com |
| L4048 GB GDDR6 | $0.46spot $0.40 | $0.19 | −59% | $0.0953% off ours | getdeploying.com |
| RTX 509032 GB GDDR7 | $0.40spot $0.15 | $0.19 | −53% | $0.0953% off ours | getdeploying.com |
| RTX 6000 Ada48 GB GDDR6 | $0.53spot $0.27 | $0.19 | −64% | $0.0953% off ours | getdeploying.com |
| RTX PRO 4500 Blackwell32 GB GDDR7 | $0.29spot $0.19 | $0.19 | −34% | $0.0953% off ours | getdeploying.com |
| A100 40GB40 GB HBM2 | $0.47spot $0.40 | $0.17 | −64% | $0.0947% off ours | getdeploying.com |
| RTX 5000 Ada32 GB GDDR6 | $0.64spot $0.40 | $0.17 | −73% | $0.0947% off ours | getdeploying.com |
| RTX A600048 GB GDDR6 | $0.47spot $0.33 | $0.17 | −64% | $0.0947% off ours | getdeploying.com |
| A1024 GB GDDR6 | $0.20 | $0.15 | −25% | $0.0940% off ours | getdeploying.com |
| RTX 4500 Ada24 GB GDDR6 | $0.34spot $0.14 | $0.15 | −56% | $0.0940% off ours | getdeploying.com |
| RTX PRO 4000 Blackwell24 GB GDDR7 | $0.20spot $0.11 | $0.15 | −25% | $0.0847% off ours | getdeploying.com |
| A4048 GB GDDR6 | not listed | $0.09 | — | $0.0633% off ours | console.vast.ai |
| RTX 4000 Ada20 GB GDDR6 | $0.17 | $0.09 | −48% | $0.0633% off ours | computeprices.com |
| RTX 4070 Ti SUPER16 GB GDDR6X | $0.12spot $0.10 | $0.09 | −25% | $0.0633% off ours | getdeploying.com |
| RTX 408016 GB GDDR6X | $0.14spot $0.12 | $0.09 | −36% | $0.0633% off ours | getdeploying.com |
| RTX 409024 GB GDDR6X | $0.30spot $0.15 | $0.09 | −70% | $0.0633% off ours | gpucloudcost.com |
| RTX 5070 Ti16 GB GDDR7 | $0.12spot $0.10 | $0.09 | −25% | $0.0633% off ours | getdeploying.com |
| RTX 508016 GB GDDR7 | $0.16spot $0.14 | $0.09 | −44% | $0.0633% off ours | getdeploying.com |
| RTX A500024 GB GDDR6 | $0.23spot $0.11 | $0.09 | −61% | $0.0633% off ours | getdeploying.com |
| L424 GB GDDR6 | $0.32spot $0.20 | $0.08 | −75% | $0.0538% off ours | getdeploying.com |
| RTX 309024 GB GDDR6X | $0.12spot $0.07 | $0.08 | −31% | $0.0538% off ours | gpucloudcost.com |
| RTX 4070 Ti12 GB GDDR6X | $0.10 | $0.08 | −20% | $0.0538% off ours | getdeploying.com |
| V100 16GB16 GB HBM2 | $0.10spot $0.06 | $0.08 | −20% | $0.0450% off ours | getdeploying.com |
| RTX 3080 Ti12 GB GDDR6X | $0.12spot $0.11 | $0.07 | −42% | $0.0443% off ours | getdeploying.com |
| T416 GB GDDR6 | $0.10spot $0.06 | $0.07 | −30% | $0.0443% off ours | getdeploying.com |
| RTX 308010 GB GDDR6X | $0.09spot $0.04 | $0.06 | −33% | $0.0350% off ours | getdeploying.com |
| RTX 407012 GB GDDR6X | $0.08spot $0.02 | $0.06 | −25% | $0.0433% off ours | getdeploying.com |
| RTX 30708 GB GDDR6 | $0.07spot $0.05 | $0.05 | −29% | $0.0340% off ours | getdeploying.com |
| RTX A400016 GB GDDR6 | $0.07spot $0.06 | $0.05 | −29% | $0.0340% off ours | getdeploying.com |
Rows where Vast.ai does not publish a price for that model are left blank rather than guessed. Spot prices are only shown when the provider publishes one.
What this comparison does — and does not — claim.
- What is compared
- One GPU, one hour, in US dollars, excluding storage, egress and taxes. Same GPU model on both sides, matched by the exact chip, not by marketing name.
- Where their prices come from
- getdeploying.com (28), gpucloudcost.com (5), computeprices.com (3), console.vast.ai (2), cloud-gpus.com (1) — the link is on every row. Recorded 15 September 2026.
- What is not compared
- Capacity, regions, network, support, interruption behaviour, billing granularity, egress fees, minimum commitments. We do not measure those on their side, so we do not assert anything about them.
- What we commit to on our side
- Per-minute billing, a 2-minute notice before a spot reclaim, disks that survive it, no egress fees, no minimum spend. All of it is on the pricing page.
- If a figure is wrong
- Prices change, and an index goes stale. Send us the link and we correct the row — that is cheaper for us than being wrong in public.
Switching, concretely
Run one job on ours
Add $40 of credit, launch the same workload, compare the invoice and the behaviour. Nothing renews.
Move the data, not the setup
No egress fees on our side; templates or your own image; a persistent disk keeps checkpoints between runs.
Keep both for a while
There is no commitment here, so there is no reason to cut anything over before you are satisfied.
About this comparison
Is this comparison up to date?
The prices attributed to Vast.ai were recorded on 15 September 2026 from the source linked on each row. Published prices move; check theirs before you switch, and tell us if a figure here is wrong — we correct it.
Are you comparing the same thing?
Same GPU model, same unit: one GPU for one hour, in US dollars, excluding storage and taxes. What differs is everything we do not measure here — capacity, regions, support, network, how an interruption is handled. This table is about price, and says nothing else.
Why is your price lower?
We buy installed capacity rather than build it, we have no sales team, and every tier of every model is priced at or below 80% of the lowest published price we can find across 137 providers. The rule is enforced by the script that publishes the catalogue. How it is checked.
What do I lose by paying less?
On the spot tier: the guarantee of keeping the GPU. We can reclaim it with a 2-minute notice, and we publish the reclaim rate for every model. Our on-demand tier is never reclaimed and is still below the cheapest on-demand price we found.
Other comparisons: every provider we track · GPU against GPU
The cheapest way to check a price claim is to run one job.
Credit from $40, billed per minute, stop whenever you want.
Billed per minute from the moment the instance is reachable. Minimum credit $40, no subscription.