🖥️ GPU Price Index

Median hourly rental prices per GPU model — compiled from live marketplace offers.

Snapshot: 2026-10-07 10:00 UTC · source: Vast.ai (public offers API) · auto-refreshed 2x daily · trends over 124 snapshots since Aug 08 (60d window)

GPU Model	Δ (60d)	Offers	Median $/hr	Min $/hr	dlperf/$
H200	▼11.2%	1	$2.633	$2.633	145
H100 NVL	▲+70.0%	2	$2.269	$2.269	167
RTX PRO 6000 Max-Q	▲+15.7%	2	$1.360	$1.281	320
RTX 6000D	▼0.0%	1	$1.257	$1.257	366
RTX PRO 6000 WS	▲+26.0%	2	$1.229	$1.122	232
RTX PRO 6000 S	▲+0.2%	7	$1.204	$0.961	436
RTX PRO 5000	▲+8.3%	3	$0.712	$0.616	310
RTX 6000Ada	▲+57.0%	1	$0.539	$0.539	210
RTX 5090	▲+39.6%	15	$0.470	$0.213	479
A100 PCIE	▲+75.5%	1	$0.470	$0.470	271
RTX 5080	▲+14.7%	2	$0.154	$0.135	795
RTX 5070 Ti	▲+1.2%	3	$0.122	$0.101	621
RTX 4090	▼33.1%	3	$0.121	$0.121	666
RTX 3090	▲+592.3%	2	$0.108	$0.088	579
RTX 4090D	▼33.2%	5	$0.107	$0.107	1,352
RTX 5070	▲+2.3%	5	$0.101	$0.096	926
RTX 3080	▲+93.6%	2	$0.082	$0.081	585
RTX 5060 Ti	▲+23.7%	1	$0.081	$0.081	1,554
RTX 4070S Ti	▼9.1%	1	$0.068	$0.068	1,199
Tesla V100	▲+148.0%	5	$0.067	$0.042	759

📌 How to read this

Prices are median asking rents per GPU-hour on Vast.ai's on-demand market — actual market clearing, not list prices.
Hot color = median. Δ = median % change over that GPU's full recorded history (▼ falling, ▲ rising); new = only one snapshot so far, — = no prior price to compare.
Sparklines show the full recorded history for each GPU — dashed grid lines and the baseline mark the price range, and the dot marks the latest price. Green = falling, red = rising.
dlperf/$ = Vast's deep learning performance score per dollar — higher = better compute value.
Low offer counts are excluded for statistical sanity.

Best $/performance today: Tesla V100 at $0.067/hr (~0 dlperf per $).

← back to imagine.josephwwu.com