Stanford University
Stanford Data Science
Marlowe Marlowe
  • Overview
  • Research Focus
  • Updates
  • Documentation
  • Apply for Access

Updates

News, milestones, and system updates from the Marlowe team.

  • Feb 20, 2026
    maintenance

    Planned Maintenance Window: February 24-25

    Marlowe will undergo scheduled maintenance on February 24-25 for firmware updates and network fabric optimization. The maintenance window is expected to last approximately 8 hours. Running jobs will be checkpointed where possible, and the scheduling queue will drain starting February...

  • Feb 15, 2026
    talks seminars

    GPU Hackathon Series Announced with Doerr School

    In partnership with the Doerr School of Sustainability, Marlowe is launching a GPU hackathon series designed to help research groups go from zero to multi-node training. The hands-on workshops will provide dedicated Marlowe time, expert guidance from Research Data Scientists,...

  • Feb 10, 2026
    talks seminars

    Marlowe Seminar Series: Scaling Research with GPUs

    The Marlowe team hosted the first event in a new seminar series focused on GPU-accelerated research at Stanford. Presentations covered multi-node training best practices, InfiniBand optimization, and case studies from Cycle 4 researchers who scaled their workflows from single-GPU to...

  • Feb 7, 2026
    research milestone

    Milestone: 80-Billion-Parameter Brain Model Validated on Marlowe

    Professor Dan Yamins’ team completed a proof-of-concept hero run on Marlowe, validating that an 80-billion-parameter brain-inspired neural network can train at scale across 24 nodes (192 H100 GPUs). The 24-hour run achieved 45% model FLOPS utilization (MFU), confirming that capability-scale...

  • Feb 6, 2026
    system status

    Cycle 5 Allocations Begin

    Marlowe’s fifth allocation cycle has officially launched, with allocations awarded to research groups across seven Stanford schools. Cycle 5 represents the largest allocation round to date, with multiple groups receiving 8+ node allocations for foundation model training. New in Cycle...

  • Dec 15, 2024
    press

    Stanford Welcomes First GPU-Based Supercomputer

    Stanford Report published a feature story on Marlowe, Stanford’s first GPU-based supercomputer. The article covers Marlowe’s role in democratizing frontier AI research at the university, making capabilities previously available only to industry accessible to faculty across seven schools. Read the...

Marlowe

Stanford's GPU-based computational instrument

Quick Links

  • Apply for Access
  • Application Guide
  • Documentation

Resources

  • Recharge Rates
  • Cite Marlowe
Stanford Research, Stanford Data Science, Stanford University IT Research Computing

© Stanford University. Stanford, California 94305.