- Cluster-first job submission:
tp job push {tp_config} {cluster_id} - New job
--teardownflag to auto-destroy clusters after job completion - New
tp cluster list --instancesflag to list all instances - Improved REST endpoints for async operation (cluster create/destroy, storage create/destroy/attach/detach, job cancel)
- New request API endpoint:
GET /request/info/{request_id}withRetry-Afterheaders
1/21/2026
- Multi-node clusters now come with Slurm preinstalled
- Shared storage can now be attached to CPU instances
1/20/2026
- Add 32xCPU instance type
1/9/2026
- Add 64xCPU instance type
12/14/2025
- Add cluster operations log to cluster information on web dashboard
- Flex storage volumes
tp storage attach <storage_id> <cluster_id>is nowtp cluster attach <cluster_id> <storage_id>tp storage detach <storage_id> <cluster_id>is nowtp cluster detach <cluster_id> <storage_id>
11/27/2025
- Introduce storage statuses
- Asynchronous cluster operations by default (cluster create, cluster destroy, storage create, storage destroy)
- Reliability improvements for jobs
- Reliability improvements for operations done on multi node clusters
tp me preferencescommandtp ssh keycommands are now undertp me sshkeytp nfscommands are now undertp storage- The
tp ssh connect <instance_id>is nowtp ssh <instance_id>
11/12/2025
- Asynchronous job cancellation by default
11/6/2025
- Cluster create & destroy email preferences
10/27/2025
- Fix non zero exit code issue on job push with certain websocket close codes
- Job push improve job output logs
- Job pull improve logging
- Change from IDs from snowflake to nanoid for readability
10/25/2025
- Faster job output uploads
- Fixes for job log streaming with carriage returns
- Fixed issues with large job outputs
- Job log stream listen improvements
- Job pull improvements for running jobs
- Created changelog
- Multi-node H200 & B200 clusters
- Jobs revamped
- SSH Key management