""" GPU Instance Library — per-GPU on-demand hourly pricing for inference Pricing as of March 31, 2026 Sources: aws.amazon.com, cloud.google.com, azure.microsoft.com, coreweave.com, crusoe.ai, fluidstack.io, lambda.ai, runpod.io, together.ai, vast.ai """ GPU_LIBRARY = { # ── AWS ────────────────────────────────────────────────────────────────── "AWS - A100 40GB - p4d.24xlarge": {"provider": "AWS", "gpu": "A100", "cost_hr": 2.75, "vram_gb": 40, "notes": "8-GPU node, ~$21.96/hr total"}, "AWS - B200 - p6-b200.48xlarge": {"provider": "AWS", "gpu": "B200", "cost_hr": 14.24, "vram_gb": 192, "notes": "8-GPU node, ~$113.93/hr total"}, "AWS - H100 SXM - p5.48xlarge": {"provider": "AWS", "gpu": "H100", "cost_hr": 6.88, "vram_gb": 80, "notes": "8-GPU node, ~$55.04/hr total"}, "AWS - H200 - p5en.48xlarge": {"provider": "AWS", "gpu": "H200", "cost_hr": 7.91, "vram_gb": 141, "notes": "8-GPU node, ~$63.30/hr total"}, "AWS - L4 - g6.xlarge": {"provider": "AWS", "gpu": "L4", "cost_hr": 0.81, "vram_gb": 24, "notes": "1-GPU instance"}, "AWS - L4 - g6.12xlarge": {"provider": "AWS", "gpu": "L4", "cost_hr": 1.15, "vram_gb": 24, "notes": "4-GPU node, ~$4.60/hr total"}, "AWS - L40S - g6e.xlarge": {"provider": "AWS", "gpu": "L40S", "cost_hr": 1.86, "vram_gb": 48, "notes": "1-GPU instance"}, "AWS - L40S - g6e.12xlarge": {"provider": "AWS", "gpu": "L40S", "cost_hr": 2.62, "vram_gb": 48, "notes": "4-GPU node, ~$10.49/hr total"}, # ── Azure ──────────────────────────────────────────────────────────────── "Azure - A100 80GB - NC A100 v4": {"provider": "Azure", "gpu": "A100", "cost_hr": 3.67, "vram_gb": 80, "notes": "1-GPU instance"}, "Azure - H100 - ND H100 v5": {"provider": "Azure", "gpu": "H100", "cost_hr": 12.29, "vram_gb": 80, "notes": "8-GPU node, ~$98.32/hr total"}, "Azure - H200 - ND H200 v5": {"provider": "Azure", "gpu": "H200", "cost_hr": 13.78, "vram_gb": 141, "notes": "8-GPU node, ~$110.24/hr total"}, "Azure - MI300X - ND MI300X v5": {"provider": "Azure", "gpu": "MI300X", "cost_hr": 7.86, "vram_gb": 192, "notes": "8-GPU node, ~$62.85/hr total"}, # ── CoreWeave ──────────────────────────────────────────────────────────── "CoreWeave - A100 80GB": {"provider": "CoreWeave", "gpu": "A100", "cost_hr": 2.70, "vram_gb": 80, "notes": "8-GPU node, ~$21.60/hr total"}, "CoreWeave - B200": {"provider": "CoreWeave", "gpu": "B200", "cost_hr": 8.60, "vram_gb": 192, "notes": "8-GPU node, ~$68.80/hr total"}, "CoreWeave - GB200 NVL72": {"provider": "CoreWeave", "gpu": "GB200", "cost_hr": 10.50, "vram_gb": 186, "notes": "4-GPU node, ~$42.00/hr total"}, "CoreWeave - H100 SXM": {"provider": "CoreWeave", "gpu": "H100", "cost_hr": 6.16, "vram_gb": 80, "notes": "8-GPU node, ~$49.24/hr total"}, "CoreWeave - H200": {"provider": "CoreWeave", "gpu": "H200", "cost_hr": 6.31, "vram_gb": 141, "notes": "8-GPU node, ~$50.44/hr total"}, "CoreWeave - L40S": {"provider": "CoreWeave", "gpu": "L40S", "cost_hr": 2.25, "vram_gb": 48, "notes": "8-GPU node, ~$18.00/hr total"}, # ── Crusoe ─────────────────────────────────────────────────────────────── "Crusoe - A100 PCIe 40GB": {"provider": "Crusoe", "gpu": "A100", "cost_hr": 1.45, "vram_gb": 40, "notes": "On-demand"}, "Crusoe - A100 SXM 80GB": {"provider": "Crusoe", "gpu": "A100", "cost_hr": 1.95, "vram_gb": 80, "notes": "On-demand"}, "Crusoe - A100 PCIe 80GB": {"provider": "Crusoe", "gpu": "A100", "cost_hr": 1.65, "vram_gb": 80, "notes": "On-demand"}, "Crusoe - H100 HGX": {"provider": "Crusoe", "gpu": "H100", "cost_hr": 3.90, "vram_gb": 80, "notes": "On-demand"}, "Crusoe - H200 HGX": {"provider": "Crusoe", "gpu": "H200", "cost_hr": 4.29, "vram_gb": 141, "notes": "On-demand"}, "Crusoe - L40S": {"provider": "Crusoe", "gpu": "L40S", "cost_hr": 1.00, "vram_gb": 48, "notes": "On-demand"}, "Crusoe - MI300X": {"provider": "Crusoe", "gpu": "MI300X", "cost_hr": 3.45, "vram_gb": 192, "notes": "On-demand"}, # ── FluidStack ─────────────────────────────────────────────────────────── "FluidStack - A100 SXM 80GB": {"provider": "FluidStack", "gpu": "A100", "cost_hr": 1.30, "vram_gb": 80, "notes": "On-demand"}, "FluidStack - H100 SXM": {"provider": "FluidStack", "gpu": "H100", "cost_hr": 2.10, "vram_gb": 80, "notes": "On-demand"}, "FluidStack - H200": {"provider": "FluidStack", "gpu": "H200", "cost_hr": 2.30, "vram_gb": 141, "notes": "On-demand"}, "FluidStack - L40S": {"provider": "FluidStack", "gpu": "L40S", "cost_hr": 1.30, "vram_gb": 48, "notes": "On-demand"}, # ── GCP ────────────────────────────────────────────────────────────────── "GCP - A100 40GB - a2-highgpu-1g": {"provider": "GCP", "gpu": "A100", "cost_hr": 3.67, "vram_gb": 40, "notes": "1-GPU instance"}, "GCP - H100 SXM - a3-highgpu-8g": {"provider": "GCP", "gpu": "H100", "cost_hr": 10.98, "vram_gb": 80, "notes": "8-GPU node, ~$87.83/hr total"}, "GCP - H200 - a3-ultragpu-8g": {"provider": "GCP", "gpu": "H200", "cost_hr": 10.85, "vram_gb": 141, "notes": "8-GPU node, ~$86.76/hr total"}, "GCP - L4 - g2-standard-4": {"provider": "GCP", "gpu": "L4", "cost_hr": 0.71, "vram_gb": 24, "notes": "1-GPU instance"}, "GCP - L4 - g2-standard-48": {"provider": "GCP", "gpu": "L4", "cost_hr": 1.00, "vram_gb": 24, "notes": "4-GPU node, ~$4.00/hr total"}, # ── Lambda ─────────────────────────────────────────────────────────────── "Lambda - A100 SXM 40GB": {"provider": "Lambda", "gpu": "A100", "cost_hr": 1.48, "vram_gb": 40, "notes": "1-GPU instance"}, "Lambda - A100 SXM 80GB": {"provider": "Lambda", "gpu": "A100", "cost_hr": 2.06, "vram_gb": 80, "notes": "8-GPU node pricing"}, "Lambda - B200 SXM": {"provider": "Lambda", "gpu": "B200", "cost_hr": 5.74, "vram_gb": 192, "notes": "8-GPU node pricing"}, "Lambda - H100 PCIe": {"provider": "Lambda", "gpu": "H100", "cost_hr": 2.86, "vram_gb": 80, "notes": "1-GPU instance"}, "Lambda - H100 SXM": {"provider": "Lambda", "gpu": "H100", "cost_hr": 3.44, "vram_gb": 80, "notes": "8-GPU node pricing"}, # ── RunPod ─────────────────────────────────────────────────────────────── "RunPod - A100 PCIe 80GB": {"provider": "RunPod", "gpu": "A100", "cost_hr": 1.19, "vram_gb": 80, "notes": "Secure Cloud on-demand"}, "RunPod - A100 SXM 80GB": {"provider": "RunPod", "gpu": "A100", "cost_hr": 1.39, "vram_gb": 80, "notes": "Secure Cloud on-demand"}, "RunPod - B200": {"provider": "RunPod", "gpu": "B200", "cost_hr": 5.98, "vram_gb": 192, "notes": "Secure Cloud on-demand"}, "RunPod - H100 PCIe": {"provider": "RunPod", "gpu": "H100", "cost_hr": 1.99, "vram_gb": 80, "notes": "Secure Cloud on-demand"}, "RunPod - H100 SXM": {"provider": "RunPod", "gpu": "H100", "cost_hr": 2.69, "vram_gb": 80, "notes": "Secure Cloud on-demand"}, "RunPod - H200 SXM": {"provider": "RunPod", "gpu": "H200", "cost_hr": 3.59, "vram_gb": 141, "notes": "Secure Cloud on-demand"}, "RunPod - L4": {"provider": "RunPod", "gpu": "L4", "cost_hr": 0.44, "vram_gb": 24, "notes": "Secure Cloud on-demand"}, "RunPod - L40S": {"provider": "RunPod", "gpu": "L40S", "cost_hr": 0.79, "vram_gb": 48, "notes": "Secure Cloud on-demand"}, # ── Together AI ────────────────────────────────────────────────────────── "Together - B200": {"provider": "Together", "gpu": "B200", "cost_hr": 7.49, "vram_gb": 192, "notes": "GPU cluster on-demand"}, "Together - H100": {"provider": "Together", "gpu": "H100", "cost_hr": 3.49, "vram_gb": 80, "notes": "GPU cluster on-demand"}, "Together - H200": {"provider": "Together", "gpu": "H200", "cost_hr": 4.19, "vram_gb": 141, "notes": "GPU cluster on-demand"}, # ── Vast.ai ────────────────────────────────────────────────────────────── "Vast.ai - A100 SXM 80GB": {"provider": "Vast.ai", "gpu": "A100", "cost_hr": 0.77, "vram_gb": 80, "notes": "Marketplace median"}, "Vast.ai - B200": {"provider": "Vast.ai", "gpu": "B200", "cost_hr": 2.67, "vram_gb": 192, "notes": "Marketplace pricing"}, "Vast.ai - H100 SXM": {"provider": "Vast.ai", "gpu": "H100", "cost_hr": 1.50, "vram_gb": 80, "notes": "Marketplace median"}, "Vast.ai - H200": {"provider": "Vast.ai", "gpu": "H200", "cost_hr": 2.20, "vram_gb": 141, "notes": "Marketplace median"}, "Vast.ai - L4": {"provider": "Vast.ai", "gpu": "L4", "cost_hr": 0.34, "vram_gb": 24, "notes": "Marketplace median"}, "Vast.ai - L40S": {"provider": "Vast.ai", "gpu": "L40S", "cost_hr": 0.50, "vram_gb": 48, "notes": "Marketplace median"}, } GPU_PROVIDERS = sorted(set(v["provider"] for v in GPU_LIBRARY.values()))