InstantML
instantml.ai
- Free plan
- Web
- API
What InstantML is
InstantML is an experiment-tracking service for logging machine-learning runs, comparing metrics, inspecting artifacts, and exporting data. It works through the web and an API, with a Python SDK built around init, log, and finish calls; the quickstart requires Python 3.11 or newer. Charts support streamed scalar series, grouped averages, smoothing, zoom, and hover readouts. Configs, metrics, tags, artifacts, histograms, and media are queryable. The SDK buffers events and spools them offline when connectivity drops. Teams can import W&B, MLflow, and Neptune JSON or shadow W&B scalar logs during migration. PyTorch, JAX, and TRL are listed alongside the SDK. Artifacts and checkpoints can use customer S3 or R2 storage. The free plan includes two seats, 2 GiB hosted storage, two projects, and 100 runs. Pro costs $199 a month; Premium costs $699 a month and adds a BYOC GCP ClickHouse option.
Who it suits
InstantML suits machine-learning teams that need run comparison, artifact tracking, and API access. Its migration imports and offline logging also fit teams changing tracking systems.
Good to know, and what to watch
- Queryable configs, metrics, tags, and artifacts
- Offline SDK logging when networks drop
- Imports W&B, MLflow, and Neptune JSON
- Customer S3 or R2 artifact storage
- Free plan available
- Python quickstart requires version 3.11 or newer
- Free plan has 2 GiB storage
- API requests and storage overages are metered
- Enterprise options use custom terms
Plans
Every plan and priceFree
Pro
Premium
Free plan
The verdict
InstantML combines web, API, and Python tracking with migration and storage options. The free tier is limited, while paid plans expand seats, projects, runs, and usage allowances.
What the maker publishes
- Artifact storage
- Artifacts and checkpoints can live in the customer's S3 or R2 storage. source
- Charts
- Charts support streamed scalar series, grouped averages, smoothing, zoom, and hover readouts. source
- Enterprise options
- Enterprise/custom terms are reserved for SSO/SAML, VPC or self-hosted deployment, custom retention, compliance, dedicated support, or unusually heavy data footprints. source
- Frameworks
- The product page lists PyTorch, JAX, and TRL alongside its Python SDK. source
- Imports
- InstantML supports importing W&B, MLflow, and Neptune JSON and can shadow W&B scalar logs during migration. source
- Offline logging
- The Python SDK buffers logging and spools offline when the network drops. source
- Premium BYOC
- Premium includes a BYOC GCP ClickHouse option, and the docs say tenant product data can be routed to customer-owned self-hosted GCP ClickHouse while artifact bytes remain on InstantML-hosted object storage. source
- Product
- InstantML describes itself as experiment tracking for logging runs, comparing metrics, inspecting artifacts, and exporting data. source
- Run data
- Configs, metrics, tags, artifacts, histograms, and media are queryable. source
- SDK
- The product provides a Python SDK with init, log, and finish calls, and the docs specify Python 3.11 or newer for the quickstart. source
- Security controls
- The SDK docs describe scoped API keys, project-scoped access, and say plaintext keys are shown once while InstantML stores their hash afterward. source
- Support
- The product page says the team will help customers import a project and compare real runs, and lists [email protected] for contact. source
- Usage limits
- Free and non-billable organizations are blocked at their monthly API request allowance; paid plans meter API requests and storage overage. source
Checked against the maker's pages on 7 Oct 2026.