GPUs

NVIDIA GPUs for open model inference

The published catalog currently carries 2 measured GPUs across 4 published packages.

View Model APIs

Measured GPUs
B300, 1 measured package
B300
1 measured package
Measured models
Kimi K3: 2.12x faster
GPU time is not currently sold. Historical rates are reference values, not an offer. Active-tier rates refer to reserved deployments. No historical rate is published for this GPU.
NVIDIA, Sep 21, 2026
Memory, NVIDIA spec
288 GB HBM3e (max)
Bandwidth, NVIDIA spec
8 TB/s (per GPU)
Historical hourly reference
Hourly price not published.
Historical active-tier reference
Active hourly price not published.
H100, 3 measured packages
H100
3 measured packages
Measured models
AREX-Turbo: 1.2x faster, Qwen3.6 27B: 1.28x faster, Qwythos-9B-Claude-Mythos-5-1M: 1.29x faster
GPU time is not currently sold. Historical rates are reference values, not an offer. Active-tier rates refer to reserved deployments. Rate source: GPU rate catalog, captured Aug 9, 2026.
NVIDIA, Aug 8, 2026
Memory, NVIDIA spec
80 GB
Bandwidth, NVIDIA spec
3,350 GB/s
Historical hourly reference
$4.18 per GPU-hour
Historical active-tier reference
$3.35 per GPU-hour
Other GPUs
A100 40GB, 40 GB, 1,555 GB/s
A100 40GB
0 measured packages
Measured models
No measured models for this GPU yet.
GPU time is not currently sold. Historical rates are reference values, not an offer. Active-tier rates refer to reserved deployments. Rate source: deploy catalog, captured Aug 9, 2026.
NVIDIA, Aug 8, 2026
Memory, NVIDIA spec
40 GB
Bandwidth, NVIDIA spec
1,555 GB/s
Historical hourly reference
$2.10 per GPU-hour
Historical active-tier reference
Active hourly price not published.
A100 80GB, 80 GB, 2,039 GB/s
A100 80GB
0 measured packages
Measured models
No measured models for this GPU yet.
GPU time is not currently sold. Historical rates are reference values, not an offer. Active-tier rates refer to reserved deployments. Rate source: GPU rate catalog, captured Aug 9, 2026.
NVIDIA, Aug 8, 2026
Memory, NVIDIA spec
80 GB
Bandwidth, NVIDIA spec
2,039 GB/s
Historical hourly reference
$2.72 per GPU-hour
Historical active-tier reference
$2.17 per GPU-hour
A10G, 24 GB, 600 GB/s
A10G
0 measured packages
Measured models
No measured models for this GPU yet.
GPU time is not currently sold. Historical rates are reference values, not an offer. Active-tier rates refer to reserved deployments. Rate source: deploy catalog, captured Aug 9, 2026.
NVIDIA, Aug 8, 2026
Memory, NVIDIA spec
24 GB
Bandwidth, NVIDIA spec
600 GB/s
Historical hourly reference
$1.10 per GPU-hour
Historical active-tier reference
Active hourly price not published.
B200, 180 GB per GPU, DGX B200, 8,000 GB/s
B200
0 measured packages
Measured models
No measured models for this GPU yet.
GPU time is not currently sold. Historical rates are reference values, not an offer. Active-tier rates refer to reserved deployments. Rate source: GPU rate catalog, captured Aug 9, 2026.
NVIDIA, Aug 8, 2026
Memory per GPU, DGX B200
180 GB
Bandwidth, NVIDIA spec
8,000 GB/s
Historical hourly reference
$8.64 per GPU-hour
Historical active-tier reference
$6.84 per GPU-hour
H200, 141 GB, 4,800 GB/s
H200
0 measured packages
Measured models
No measured models for this GPU yet.
GPU time is not currently sold. Historical rates are reference values, not an offer. Active-tier rates refer to reserved deployments. Rate source: GPU rate catalog, captured Aug 9, 2026.
NVIDIA, Aug 8, 2026
Memory, NVIDIA spec
141 GB
Bandwidth, NVIDIA spec
4,800 GB/s
Historical hourly reference
$5.58 per GPU-hour
Historical active-tier reference
$4.46 per GPU-hour
L4, 24 GB, 300 GB/s
L4
0 measured packages
Measured models
No measured models for this GPU yet.
GPU time is not currently sold. Historical rates are reference values, not an offer. Active-tier rates refer to reserved deployments. Rate source: GPU rate catalog, captured Aug 9, 2026.
NVIDIA, Aug 8, 2026
Memory, NVIDIA spec
24 GB
Bandwidth, NVIDIA spec
300 GB/s
Historical hourly reference
$0.69 per GPU-hour
Historical active-tier reference
$0.48 per GPU-hour
L40S, 48 GB, 864 GB/s
L40S
0 measured packages
Measured models
No measured models for this GPU yet.
GPU time is not currently sold. Historical rates are reference values, not an offer. Active-tier rates refer to reserved deployments. Rate source: GPU rate catalog, captured Aug 9, 2026.
NVIDIA, Aug 8, 2026
Memory, NVIDIA spec
48 GB
Bandwidth, NVIDIA spec
864 GB/s
Historical hourly reference
$1.90 per GPU-hour
Historical active-tier reference
$1.33 per GPU-hour
T4, 16 GB, 320 GB/s
T4
0 measured packages
Measured models
No measured models for this GPU yet.
GPU time is not currently sold. Historical rates are reference values, not an offer. Active-tier rates refer to reserved deployments. Rate source: deploy catalog, captured Aug 9, 2026.
NVIDIA, Aug 8, 2026
Memory, NVIDIA spec
16 GB
Bandwidth, NVIDIA spec
320 GB/s
Historical hourly reference
$0.59 per GPU-hour
Historical active-tier reference
Active hourly price not published.

Call Nemotron 3.5 Lightning 30B with Model APIs

Use a workspace API key and pay for input, cached input, and output tokens.

View Model APIs

TLS in transit, AES-256 at rest

Workspace isolation

No training on your data

SOC 2 Type II

Measured GPUs

B300, 1 measured package

B300

1 measured package

Measured models

Kimi K3: 2.12x faster

GPU time is not currently sold. Historical rates are reference values, not an offer. Active-tier rates refer to reserved deployments. No historical rate is published for this GPU.

NVIDIA, Sep 21, 2026

Memory, NVIDIA spec: 288 GB HBM3e (max)
Bandwidth, NVIDIA spec: 8 TB/s (per GPU)
Historical hourly reference: Hourly price not published.
Historical active-tier reference: Active hourly price not published.

Other GPUs

A100 40GB, 40 GB, 1,555 GB/s

A100 40GB

0 measured packages

Measured models

No measured models for this GPU yet.

GPU time is not currently sold. Historical rates are reference values, not an offer. Active-tier rates refer to reserved deployments. Rate source: deploy catalog, captured Aug 9, 2026.

NVIDIA, Aug 8, 2026

Memory, NVIDIA spec: 40 GB
Bandwidth, NVIDIA spec: 1,555 GB/s
Historical hourly reference: $2.10 per GPU-hour
Historical active-tier reference: Active hourly price not published.