Skip to content
TheChatArchiveThe Chat Archive
Perplexitysonar-reasoning-pro-high· PerplexityHugging Face LMArena Search Arena 24K

looking for best cloud gpu offers with free trials, make a list of at least 20 offers…

From a Chatbot Arena battleThe same person sent this prompt to two anonymous models at once. Compare it with Geminigemini-2.5-flash-preview-04-17-grounding’s answer.
  1. User

    looking for best cloud gpu offers with free trials, make a list of at least 20 offers including terms, credit, hourly cost, credit card requriment

  2. Perplexity Assistant

    Comprehensive List of Cloud GPU Providers with Free Trials and Credits

    Here's a detailed list of cloud GPU services offering free trials, credits, or free tiers, including their terms and pricing information:

    Major Cloud Providers

    1. Google Colab (Free)

      • Free GPU: Nvidia K80s or Tesla T4 (16GB memory)
      • Terms: 12 hours of free GPU access
      • Idle timeout: 30 minutes
      • Credit card required: No
      • Limitations: Notebooks only, shared GPU[1][5]
    2. Google Cloud Platform

      • Free credit: $300 valid for 90 days
      • GPU pricing: T4 at 0.35/hour,V100at0.35/hour, V100 at 2.48/hour
      • Credit card required: Yes
      • Note: Credits typically used within days for GPU workloads[1][7][8][9]
    3. Amazon Sagemaker Studio Lab

      • Free GPU: Tesla T4 (16GB memory, not shared)
      • Storage: 15GB persistent
      • Terms: 4 hours runtime, 8 hours per day maximum
      • Credit card required: No
      • Note: Requires application approval, background executions possible[1][9]
    4. Amazon AWS

      • Free tier: 750 hours of t2.micro + startup credits
      • GPU pricing: g4dn.xlarge at $0.53/hour
      • Credit card required: Yes
      • Note: Good for lightweight workloads[9]

    Specialized GPU Platforms

    1. Codesphere

      • Free GPU: Shared GPU with signup
      • Storage: 20GB
      • Idle timeout: 60 minutes
      • Credit card required: No
      • Features: Unlimited monthly hours, not restricted to notebooks[1]
    2. Gradient by Paperspace

      • Free tier: Highest RAM offering (30GB)
      • Storage: 5GB
      • Idle time: 1-6 hours
      • Session limit: 6 hours maximum
      • Credit card required: Not specified
      • Limitations: Notebooks only, all notebooks are public[1][6]
    3. Thunder Compute

      • Free credit: $20 monthly
      • GPU pricing: T4 at 0.27/hour,A100at0.27/hour, A100 at 0.57/hour
      • Credit card required: Not specified
      • Note: Up to 80% cheaper than AWS/GCP[9]
    4. Paperspace CORE

      • Free credit: $10 of free Gradient credit for fast.ai users
      • GPU pricing: H100 80GB at 5.95/hour,RTXA6000at5.95/hour, RTX A6000 at 1.89/hour
      • Credit card required: Not specified
      • Note: Owned by DigitalOcean[6]
    5. Vast.ai

      • Pricing model: Marketplace for GPU rentals
      • Median cost: $0.15/hour
      • Credit card required: Not specified
      • Tip: Uncheck "verified machines" filter for cheaper rates[3][9]
    6. Kaggle

      • Free GPU access with account creation
      • Terms: Limited session time
      • Credit card required: No
      • Features: Good file system and secrets management[1][5]
    7. GPU Mart

      • Free GPU trial available
      • GPU models: 20+ NVIDIA GPU servers
      • Credit card required: Not specified
      • Details: Limited information available[4]

    Cost-Effective Options

    1. TensorDock

      • Pricing: From $0.05/hour (spot instances)
      • High-end GPUs: H100 SXM at $2.25/hour
      • Credit card required: Not specified
      • Note: One of the cheapest options for spot instances[9]
    2. RunPod

      • Pricing: A40 from 0.44/hour,A10080GB(community)at0.44/hour, A100 80GB (community) at 1.19/hour
      • Billing: Per minute
      • Credit card required: Not specified
      • Features: Quick setup, 50+ pre-configured templates[9][10]
    3. DataCrunch

      • Pricing: V100 at 0.39/hour,H100at0.39/hour, H100 at 3.35/hour
      • Credit card required: Not specified
      • Note: Up to 8x cheaper than major cloud providers[6]
    4. OVHcloud

      • Pricing: H100 80GB at 3.39/hour,A10080GBat3.39/hour, A100 80GB at 3.35/hour
      • Credit card required: Not specified
      • Note: Europe's largest cloud provider[6]
    5. Scaleway

      • Pricing: H100 80GB at 2.73/hour,L40S48GBat2.73/hour, L40S 48GB at 1.40/hour
      • Credit card required: Not specified
      • Note: One of the lowest H100 hourly rates among large providers[6]
    6. Lambda Labs

      • Pricing: From $0.50/hour
      • Credit card required: Not specified
      • Note: Popular among ML researchers[9]

    Special Programs

    1. AWS Cloud Credits for Research

      • Free research credits available through application
      • Credit card required: No for approved researchers
      • Note: Application-based approval process[2]
    2. Nvidia GPU Grant Program

      • Free GPU access for qualifying research projects
      • Credit card required: No for approved researchers
      • Note: Application-based approval process[2]
    3. Google Startup Program

      • Free credits: 1,000to1,000 to 100,000
      • Credit card required: Varies by program tier
      • Eligibility: For qualifying startups
      • Note: Application-based approval process[2]
    4. GitHub Education Pack

      • Free credits: Up to $110 for AWS
      • Credit card required: No for approved students
      • Eligibility: For verified students
      • Note: Includes many other developer tools and services[2]
    5. Google Cloud Education Grant

      • Free credits for educational purposes
      • Credit card required: No for approved educators/students
      • Note: Application-based approval process[2]

    Many of these providers offer additional discounts for students, researchers, or through special programs. For time-sensitive workloads, consider spot/preemptible instances which can reduce costs by 60-91% with some risk of interruption.

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.