Dataset Viewer
Auto-converted to Parquet Duplicate
as_of
stringdate
2026-08-03 00:00:00
2026-08-03 00:00:00
gpu
stringclasses
5 values
vast_name
stringclasses
5 values
tier
stringclasses
2 values
usd_per_gpu_hr_min
float64
0.13
6.25
usd_per_gpu_hr_p25
float64
0.47
6.25
usd_per_gpu_hr_median
float64
0.67
6.25
sample_count
int64
3
61
2026-08-03
H100
H100 SXM
on-demand
0.99
2.095
2.801
31
2026-08-03
H100
H100 SXM
interruptible
0.447
0.905
2.035
20
2026-08-03
H200
H200
on-demand
3.817
4.081
4.178
12
2026-08-03
H200
H200
interruptible
3.027
3.786
3.882
11
2026-08-03
B200
B200
on-demand
5.001
5.878
5.939
27
2026-08-03
B300
B300
on-demand
6.25
6.25
6.251
3
2026-08-03
B300
B300
interruptible
3.751
4.688
5.625
3
2026-08-03
A100
A100 SXM4
on-demand
0.268
0.747
0.935
61
2026-08-03
A100
A100 SXM4
interruptible
0.131
0.468
0.668
36

GPU spot-rental prices

Weekly medians of vast.ai marketplace asks, normalised to USD per GPU-hour (dph_total / num_gpus), for H100, H200, B200, B300, A100 — on-demand and interruptible tiers where listed. Method: cheapest-500 asks per SKU, so the series is deliberately low-biased; min / p25 / median per tier.

Machine-generated from vast.ai's public marketplace API. No internal figures, no vendor quotes.

  • data/prices.csv — the series, one row per GPU x tier per collection date (starts 2026-08-03).
  • collect.mjs — the collector; run node collect.mjs to extend the series yourself.

Maintained by thsysmfh; the same feed drives the public calculators at tensorforge.cc.

Downloads last month
55