Usage-based pay-as-you-go pricing with hourly compute charges; committed contracts available for volume discounts. Hosted option with Anyscale-managed infrastructure or Bring Your Own Cloud deployment available.
Not given
Key features
Multimodal data curation at scale
Distributed model training with elastic scaling
Batch embedding generation at scale
Multi-cloud deployment and orchestration
Cluster-backed development environments
Workload-specific observability and debugging
Production-grade managed Ray clusters
Access control and governance
GPU budget controls and cost attribution
Post-training with SkyRL and veRL support
Harpe: chooses what an AI agent reads under a token budget with consistent results
Kibisis: keeps agent memory on your hardware with AES-256-GCM encryption
Argus: append-only record that detects edits with cryptographic verification
No network code by default for local and disconnected use
MIT License with published code, benchmarks and raw results
What makes it different
Built by creators of Ray, the world's most widely adopted AI compute engine
Multi-cloud deployment without code changes
Feels local but runs distributed
Unified GPU pooling across clouds and regions
Measured empirical validation across 6,111 questions outperforming full-context models
Zero plaintext on disk with encrypted vaults matching plaintext accuracy
78.3% fewer tokens sent to the model with a 1,200-token budget per turn
Preserves conversational chronology and relational qualifiers under token budgets
Perseus Computing