RunPod GPU Stock Check

Check RunPod GPU availability by datacenter and suggest GPUs matching requirements.

Prerequisites

Set RunPod API key via either:

•Environment variable RUNPOD_API_KEY
•~/.runpod/config.toml (set via runpod CLI)

Script

${CLAUDE_PLUGIN_ROOT}/skills/stocks/fetch_gpu_stocks.py

Usage

bash

# All datacenter × GPU availability
uv run --script fetch_gpu_stocks.py

# Filter by memory (80GB+)
uv run --script fetch_gpu_stocks.py --min-memory 80

# Filter by GPU name (partial match, multiple allowed)
uv run --script fetch_gpu_stocks.py --gpu h100 5090

# Network Volume (S3) supported datacenters only
uv run --script fetch_gpu_stocks.py --storage

# Filter by stock level
uv run --script fetch_gpu_stocks.py --stock high    # High only
uv run --script fetch_gpu_stocks.py --stock medium  # High + Medium

# Filter by GPU generation
uv run --script fetch_gpu_stocks.py --gen hopper    # H100, H200
uv run --script fetch_gpu_stocks.py --gen blackwell # B200, RTX 5090, etc.
uv run --script fetch_gpu_stocks.py --gen ada       # RTX 4090, L40, etc.
uv run --script fetch_gpu_stocks.py --gen ampere    # A100, RTX 3090, etc.

# Combined filters
uv run --script fetch_gpu_stocks.py --min-memory 80 --storage --stock high

# JSON output
uv run --script fetch_gpu_stocks.py --json

Output Example

code

Datacenter   Location        GPU                          Mem Gen        Stock Storage    Price
---------------------------------------------------------------------------------------------------------
EU-CZ-1      Europe          RTX 5090                    32GB blackwell   High       ✓    $0.69
EU-RO-1      Europe          RTX 5090                    32GB blackwell   High       ✓    $0.69
US-CA-2      United States   RTX 5090                    32GB blackwell   High       ✓    $0.69
US-NC-1      United States   RTX 5090                    32GB blackwell   High       ✓    $0.69

Found 4 options across 4 datacenters, 1 GPU types

Filter Options

Option	Description
`--min-memory GB`	Minimum VRAM (GB)
`--gpu KEYWORD...`	Filter by GPU name (partial match, multiple)
`--storage`	Network Volume supported only
`--stock {high,medium,low}`	Stock level
`--gen {blackwell,hopper,ada,ampere,volta,amd}`	GPU generation
`--secure-cloud`	Secure Cloud only
`--community-cloud`	Community Cloud only
`--json`	JSON output

GPU Generation Classification

Generation	GPUs
blackwell	B200, B300, RTX PRO 6000, RTX 5090, RTX 5080
hopper	H100, H200
ada	RTX 4090, RTX 4080, L40, L40S, RTX 6000 Ada, etc.
ampere	A100, A40, RTX 3090, RTX A6000, L4
volta	V100
amd	MI300X

Examples

User: "Where can I use RTX 5090 with Network Volume?"

bash

uv run --script fetch_gpu_stocks.py --gpu 5090 --storage

User: "Which 80GB+ GPUs are in stock?"

bash

uv run --script fetch_gpu_stocks.py --min-memory 80 --stock high

User: "Hopper generation with Storage support"

bash

uv run --script fetch_gpu_stocks.py --gen hopper --storage

User: "Compare H100 and A100 availability"

bash

uv run --script fetch_gpu_stocks.py --gpu h100 a100 --storage

Important Notes

•stockStatus is unreliable: API often returns "High" regardless of actual availability
•Always verify actual availability on RunPod Web UI
•Prices shown are Community Cloud (falls back to Secure Cloud if unavailable)

User communication: When returning results, inform users that stockStatus is from the API and may not reflect actual availability. Recommend checking the Web UI for real-time status.