RunPod "exploit" steps:

Create RunPod API key (https://www.runpod.io/console/user/settings), save it somewhere.
Open up your favorite HTTP client such as Postman (https://www.postman.com/downloads) or Bruno (https://github.com/usebruno/bruno/releases)
Create new GraphQL request with the following URL: "https://api.runpod.io/graphql?api_key=YOUR_API_KEY"

Enter the following query;

⎗
✓
mutation { 
    podRentInterruptable(input: {  
        bidPerGpu: 0.17,
        cloudType: SECURE, 
        gpuCount: 1,
        volumeInGb: 80,
        containerDiskInGb: 10, 
        minVcpuCount: 2,
        minMemoryInGb: 8,
        gpuTypeId: "NVIDIA RTX A6000",
        name: "RunPod TheBloke LLMs",
        imageName: "thebloke/cuda11.8.0-ubuntu22.04-oneclick:latest",
        dockerArgs: "",
        ports: "7860/http,5000/http,22/tcp,5005/tcp",
        volumeMountPath: "/workspace",
        env: [{ key: "UI_ARGS", value: "--verbose" }]
    }) { 
        id imageName env machineId machine { podHostId costPerHr  currentPricePerGpu diskMBps location gpuAvailable maxDownloadSpeedMbps maxUploadSpeedMbps note } lowestBidPriceToResume
    } 
}

Change the image name, env variables, gpu count etc. as you wish, press send and you have your $0.17/hr A6000 (excluding storage cost).

Some additional info:

As far as I know, there are only two A6000 GPU clusters that has this low bidding, one of them is in USA other is Aussie, in the past their internet was so dogshit you had to wait 2-3 hour to install 70B model, not sure if they fixed that. Also, I dont think RunPod isnt gonna fix this as this has existed ever since they removed bidding from the frontend, this is the most cost effective way to get GPUs. Both clusters are almost always available. The image is probably outdated so go find yourself a new one. Other GPUs might have similar cost effective prices, you can find them by simply changing the gpuTypeId.