Oblivus GPU Cloud Pricing: Complete Guide ($/hr for Every GPU)

Deploybase · July 15, 2025 · GPU Pricing

Contents

Oblivus Platform Overview

Oblivus operates a GPU cloud platform providing dedicated hardware for machine learning, rendering, and computational workloads. The platform offers transparent hourly pricing with no hidden fees or long-term commitments required.

Oblivus differentiates through customer support focus. Every customer receives dedicated account manager for tier 2+ deployments. This service-first approach appeals to teams prioritizing reliability over lowest pricing.

The platform spans 8 global data centers covering Americas, Europe, and Asia-Pacific regions. Regional pricing varies 10-15% based on local demand.

Complete Pricing Table

Entry-Level GPUs (Development & Testing)

GPUVRAMPrice/HourMonthly (730h)Annual (8,760h)
RTX 409024GB$0.64$467$5,607
RTX A600048GB$0.55$402$4,818
L4048GB$1.05$767$9,198

High-Performance GPUs (Training & Large Inference)

GPUVRAMPrice/HourMonthly (730h)Annual (8,760h)
A100 PCIe (80GB)80GB$1.47$1,073$12,877
A100 NVLink (80GB)80GB$1.57$1,146$13,753
H100 PCIe80GB$1.98$1,445$17,345
H100 NVLink80GB$2.08$1,518$18,221
H100 SXM80GB$2.94$2,146$25,751

Next-Generation GPUs (Latest Performance)

GPUVRAMPrice/HourMonthly (730h)Annual (8,760h)
H200 SXM141GB$3.99$2,913$34,952

GPU Selection Guide

For Development & Testing

Choose RTX 4090 or RTX A5000 for prototyping. 24GB VRAM accommodates models up to 13B parameters with quantization.

Budget estimate: $0.58-0.85 per hour supports 8-12 hours daily experimentation at $14-25 monthly cost.

For Small Model Inference

L40S provides optimal price-to-performance for models under 30B parameters. 48GB VRAM handles quantized 70B models at acceptable latency.

Oblivus L40 pricing at $1.05/hour is competitive for inference workloads, though Oblivus does not currently list an L40S in their catalog. Compare with RunPod L40S at $0.79/hour and Lambda Labs at $0.92/hour.

For Large Model Inference

H100 SXM at $2.94/hour delivers best throughput for 70B+ parameter models. Dual H100 configuration ($5.88/hour combined) handles 140B parameter models at 20-30ms latency per token.

For Model Training

H100 SXM and H200 GPUs accelerate training significantly. H200's 141GB VRAM enables training 70B-parameter models with larger batch sizes than H100.

Training 70B parameter model on single H200: approximately $100-150 per training run.

Discount Programs

Commitment Discounts

Monthly prepayment: 10% discount on hourly rates

Annual prepayment: 25% discount on hourly rates

Example: H100 SXM at $2.94/hour becomes $2.21/hour with annual commitment.

Volume Discounts

1,000+ GPU-hours monthly: 5% discount

5,000+ GPU-hours monthly: 15% discount

50,000+ GPU-hours monthly: 25% discount (negotiated custom rates)

Startup Programs

Oblivus offers $5,000 monthly credits for qualifying startups. Program runs for 12 months with no commitment.

Total Cost Comparison

Scenario 1: Daily Small Model Development

Configuration: RTX 4090, 4 hours daily

Monthly cost: $0.64/hour × 4 hours × 22 days = $56.32

Annual cost: $675.84

This supports rapid prototyping with minimal investment.

Scenario 2: Production Inference Service

Configuration: Single L40, continuous operation

Monthly cost: $1.05/hour × 730 hours = $767

Annual cost: $9,198

L40 handles models up to 34B parameters at acceptable latency.

Scenario 3: Large Model Training Project

Configuration: Dual H100 SXM, 300 hours training

Total cost: $2.94/hour × 2 GPUs × 300 hours = $1,764

With annual commitment discount (25%): $1,323

Scenario 4: Production Continuous Inference

Configuration: 4xH100 SXM cluster, 24/7 operation with 20% annual discount

Monthly cost: $2.94/hour × 4 × 730 hours × 0.80 = $6,861

Annual cost: $82,330

This configuration handles 500K+ daily inference requests with low latency.

Use Case Recommendations

Best For Oblivus

Support-intensive workloads: Teams valuing hands-on support over lowest price benefit from Oblivus's account management.

Long-term deployments: Annual discounts favor continuous infrastructure. Production services planning >6 month duration save money through commitments.

Multi-GPU distributed training: Oblivus networking supports efficient distributed training across 4-8 GPUs.

Consider Alternatives

Spot/flexible workloads: Vast AI's marketplace pricing undercuts Oblivus by 20-30% for flexible timing.

One-time projects: Lambda Labs offers better on-demand pricing without commitment requirements.

Ultra high-volume: CoreWeave's bulk pricing becomes advantageous at 10,000+ monthly GPU-hours.

FAQ

Does Oblivus offer free trial? New accounts receive $50 credits. This covers approximately 78 hours of RTX 4090 usage ($0.64/hr) or 17 hours of H100 SXM compute ($2.94/hr).

What's the minimum commitment? Pay-as-you-go pricing available with hourly billing. No minimum usage required.

Does Oblivus provide GPUs in my region? Oblivus serves 8 regions: US East, US West, EU Central, EU West, Asia-Southeast, Asia-Northeast, South America, and Middle East. Regional pricing varies 10-15%.

Can I burst beyond provisioned capacity? Yes. Standard capacity provisioning available with burst instances at 50% premium pricing. Burst capacity accessible instantly without queue delays.

Does Oblivus include software/libraries? All instances launch with CUDA, cuDNN, PyTorch, and TensorFlow pre-installed. Custom environment setup available for additional fee ($50/instance).

What's the typical SLA? Oblivus guarantees 99.5% uptime. Service credits apply for exceeding monthly downtime threshold.

Compare comprehensive GPU pricing across all providers using Vast AI detailed pricing. Explore Lambda Cloud GPU pricing for alternative options. Review Jarvislabs GPU pricing for additional competitive analysis. Check the GPU pricing guide for comprehensive provider comparison.

Sources