Don't gamble with your cloud costs. Use Spare Cores!

Run every ML workflow faster and cheaper—no config wrangling required.

Resource Tracker watches your Metaflow runs, learns real CPU / RAM / GPU needs, and recommends the right instance for the next run.

PRICE/HOURS
CPU
RAM (GB)
CLOUD VENDOR
AWS logo
AWS logo
GCP logo
GCP logo
AWS logo
GCP logo
Hetzner logo
GCP logo
GCP logo
AWS logo
Hetzner logo
AWS logo
GCP logo
AWS logo
AWS logo
AWS logo
Hetzner logo
AWS logo
GCP logo
AWS logo
AWS logo
AWS logo
GCP logo
Hetzner logo
AWS logo
AWS logo
GCP logo
AWS logo
AWS logo
Hetzner logo
AWS logo
AWS logo
AWS logo
GCP logo
AWS logo
GCP logo
SERVER TYPE
t4g.micro
arm64
t4g.small
arm64
t4g.large
arm64
t4g.xlarge
arm64
m7a.medium
arm64
m7a.large
arm64
m7a.2xlarge
arm64
t3a.nano
arm64
t3a.small
arm64
t3a.medium
arm64
t3a.large
arm64
m4.large
arm64
m4.xlarge
arm64
t2.small
arm64
t2.medium
arm64
t2.large
arm64
t2.xlarge
arm64
t2.2xlarge
arm64
t4g.medium
arm64
t4g.medium
arm64
t4g.medium
arm64
t4g.medium
arm64
t4g.medium
arm64
t4g.medium
arm64
c6a.large
arm64
c6a.xlarge
arm64
c6a.2xlarge
arm64
t4g.large
arm64
t4g.medium
arm64
t4g.medium
arm64
t4g.medium
arm64
t4g.medium
arm64
c7i.large
x86_64
c7i.large
x86_64
c5n.large
arm64
e2-highcpu-32
x86_64
REGION
Asia Pacific
Jakarta
Europe
Milan
Asia Pacific
Jakarta
Asia Pacific
Jakarta
Asia Pacific
Hyderabad
Middle East
US East
Middle East
Middle East
Asia Pacific
Hyderabad
US East
Asia Pacific
Melbourne
Asia Pacific
Mumbai
Asia Pacific
Mumbai
Europe
Zurich
Asia Pacific
Mumbai
Asia Pacific
Hyderabad
Africa
Cape Town
Asia Pacific
Hong Kong
Middle East
Europe
Zurich
Asia Pacific
Hong Kong
Asia Pacific
Melbourne
Africa
Cape Town
Asia Pacific
Melbourne
Middle East
Europe
Milan
Asia Pacific
Jakarta
Europe
Zurich
Asia Pacific
Hong Kong
Asia Pacific
Tokyo
US East
Europe
Paris
Europe
Paris
Europe
Milan
us-west-4
Las Vegas
<why use="spare_cores">
Trusted by early adopters tracking and optimizing hundreds of Metaflow steps every day.
Every under‑utilized core is budget you could spend on the next experiment.
Track
<resource_tracker>
The Spare Cores Resource Tracker provides per-step CPU/GPU/RAM straight from your Metaflow runs, and recommends the right instance for your next run.
Know
<navigator>
The Spare Cores Navigator benchmarks 2,000+ instance types across clouds and recommends the best fit.
Do
<sentinel>
The Spare Cores Sentinel Access and Sentinel Pro (coming soon) add dashboards, richer reporting, alerts & within or cross‑cloud auto‑tune.
Built for bursty DS/ML pipelines—not 24 × 7 web traffic.
<product id="resource_tracker">
For ML Engineers & Data Scientists
No more tedious config tweaking.
@track_resources decorator → report and sizing advice.
Open‑source, MPL‑licensed—full control, no vendor lock‑in.
Metaflow today—other orchestration tools on the roadmap.
<product id="sentinel">
For Data‑Science & FinOp Leads
Your team is wasting time and money. See where every cloud dollar goes—then reclaim it without slowing the team.
Sentinel Access
(Join the wait-list for Sentinel Access)
360-view of every job you run in a central dashboard
Feature-rich reports, 7-day retention
Auto-tuning resource allocations
Proactive alerts (over-provision, drift, anomalies)
Free cloud service for individual users
Sentinel Pro
(Coming soon)
Everything in Sentinel Access, plus team support
Hierarchical or tag-based report breakouts
Full historical optimization and savings analysis
Central management (e.g. max GPU hours per project)
Custom data retention, SLA, SSO & RBAC
Optional on-prem or hybrid deployment
<init products="resource_tracker sentinel">
Ready to right‑size your workflows?
  1. Install the open-source Resource Tracker package
  2. Sign up for early access to a free Sentinel Access account to see 7‑day workflow history
  3. Book a 30‑min live call to share your pain points and preview Sentinel Access and Pro

              from resource_tracker import ResourceTracker

              # start trackers in the background (non-blocking)
              tracker = ResourceTracker()

              # your compute-heavy task
              train_model_or_render_video()

              # view a resource usage report at any time
              tracker.report()
              # or get resource allocation advice for future runs
              tracker.suggest()
            
<services from="spare_cores">
More ways Spare Cores helps
Spare Cores Navigator
Real‑time benchmarks for 2 000 + AWS, GCP, Azure & Hetzner instance types—LLM, DB, compression and more.
Cloud‑Optimization Consulting
Custom benchmarks, spend audits, and right‑sizing plans that unlock another 10–20 % savings.
Scalable Shiny / Streamlit Hosting
Monitoring and auto‑scaling containerized dashboards from zero to thousands of users on your cloud—or ours.
Ad-hoc AI Workload Sizing
Instantly pick the lowest-cost, right-sized infra for one-off tasks—like labeling 10 k PDFs with an LLM.
<
release_notes
>
We publish case studies, learnings from implementing Spare Cores components, and other project updates from time to time. Find some of our featured articles below: