Connected to Inveniam / MANTRA / Valdi – AI Data & Compute Fabric
Org: Horizon CapitalSigned in as w.buffet@horizon.cap

Diol Compute – Powered by Valdi.ai

On Demand GPUs

Curated GPU profiles from Valdi.ai and Core 42 for Diol workloads. Pricing is in USD, while settlement, staking, and rewards are handled in $DIOL on MANTRA in the background.

Optimized for Baseline Monitoring, stress tests, and Diol analytics blueprints.

Change the default GPU profile per asset, blueprint, or policy bundle. Diol handles routing, failover, and SLA enforcement across providers.

Flagship AI GPUs
Top-tier H-series profiles for production LLMs, large-scale training, and high-throughput inference.
Runtime billed per-second with 10-minute minimum.Multi-region; Diol routes jobs to the closest compliant region.
H200 141GB – US-West
Flagship
Core 42Core 42 partner
Frontier model training, large-context inference, RAG over massive embeddings.
56 vCPU / 448 GB
NVIDIA H200 SXM · 141 GB
2 TB NVMe
200 Gbit InfiniBand
US-West • US-Central
$3.00 / gpu-hr· min 10 min
Availability: Medium
Valdi SKU: VALDI-H200-SXM-141
Pricing in USD at workspace level; Diol settles in $DIOL on MANTRA in the background and routes workloads across Valdi and approved partners.
H100 80GB – SXM Cluster
Flagship
Core 42Core 42 partner
Production LLM training, multi-GPU fine-tuning, latency-sensitive inference.
48 vCPU / 384 GB
NVIDIA H100 SXM · 80 GB
2 TB NVMe
400 Gbit InfiniBand
US-West • EU-Central
$2.00 / gpu-hr· min 10 min
Availability: High
Valdi SKU: VALDI-H100-SXM-80
Pricing in USD at workspace level; Diol settles in $DIOL on MANTRA in the background and routes workloads across Valdi and approved partners.
H100 80GB – PCIe
Flagship
High-throughput inference and mid-size training jobs.
32 vCPU / 256 GB
NVIDIA H100 PCIe · 80 GB
1 TB NVMe
100 Gbit Ethernet
US-East • US-West • EU-West
$2.00 / gpu-hr· min 10 min
Availability: High
Valdi SKU: VALDI-H100-PCIe-80
Pricing in USD at workspace level; Diol settles in $DIOL on MANTRA in the background and routes workloads across Valdi and approved partners.
Balanced & Enterprise GPUs
A-series and L-series profiles tuned for steady, cost-efficient enterprise AI workloads.
Runtime billed per-second with 10-minute minimum.Multi-region; Diol routes jobs to the closest compliant region.
A100 80GB – General Purpose
Balanced
Core 42Core 42 partner
Training medium-size models, vector DBs, and mixed analytics workloads.
32 vCPU / 256 GB
NVIDIA A100 · 80 GB
1 TB NVMe
100 Gbit Ethernet
US-East • EU-West • AP-Southeast
$1.00 / gpu-hr· min 10 min
Availability: High
Valdi SKU: VALDI-A100-80
Pricing in USD at workspace level; Diol settles in $DIOL on MANTRA in the background and routes workloads across Valdi and approved partners.
L40S – Inference Optimized
Balanced
Image/video generation, foundation-model inference, enterprise RAG.
24 vCPU / 192 GB
NVIDIA L40S · 48 GB
1 TB NVMe
25 Gbit Ethernet
US-East • US-West • EU-Central
$0.75 / gpu-hr· min 10 min
Availability: High
Valdi SKU: VALDI-L40S-48
Pricing in USD at workspace level; Diol settles in $DIOL on MANTRA in the background and routes workloads across Valdi and approved partners.
A40 – Compute & Graphics
Balanced
Fine-tuning smaller models and high-fidelity 3D / graphics workloads.
20 vCPU / 160 GB
NVIDIA A40 · 48 GB
1 TB NVMe
25 Gbit Ethernet
US-East • EU-West
$0.55 / gpu-hr· min 10 min
Availability: Medium
Valdi SKU: VALDI-A40-48
Pricing in USD at workspace level; Diol settles in $DIOL on MANTRA in the background and routes workloads across Valdi and approved partners.
Burst & Legacy Profiles
RTX-class profiles for experimentation, CI, and legacy GPU-heavy jobs.
Runtime billed per-second with 10-minute minimum.Multi-region; Diol routes jobs to the closest compliant region.
RTX 4090 – Burst
Burst
Cost-optimized experiments, smaller LLMs, diffusion models, notebooks.
16 vCPU / 96 GB
NVIDIA RTX 4090 · 24 GB
750 GB NVMe
10 Gbit Ethernet
US-West
$0.45 / gpu-hr· min 10 min
Availability: High
Valdi SKU: VALDI-RTX4090-24
Pricing in USD at workspace level; Diol settles in $DIOL on MANTRA in the background and routes workloads across Valdi and approved partners.
RTX 3090 – Legacy Burst
Burst
Legacy workloads, GPU-rich CI, and large-scale test grids.
12 vCPU / 64 GB
NVIDIA RTX 3090 · 24 GB
500 GB NVMe
10 Gbit Ethernet
US-West • EU-East
$0.30 / gpu-hr· min 10 min
Availability: Medium
Valdi SKU: VALDI-RTX3090-24
Pricing in USD at workspace level; Diol settles in $DIOL on MANTRA in the background and routes workloads across Valdi and approved partners.
Compute provided by Valdi.ai and Core 42, orchestrated by Diol.Pricing and specs shown here are illustrative only and will be sourced from live Valdi / partner APIs in production.