commercial · v7.2 · SLA-backed
Linux Kernel OS · Enterprise Optimiser
Production-grade adaptive kernel platform for AI and quantum-signal workloads. Hardened eBPF programs, signed out-of-tree kmods, hybrid QPU bridge and the orchestration plumbing required to run inference fleets under contractual SLA.
Team
$2,400/ node / year
Up to 32 nodes. Business-hours support, 8h response SLA.
- →Distributed signal sharding
- →Hardware QPU bridge (IBM, IonQ)
- →Helm + RPM + DEB
- →9×5 support
Datacenter
$7,800/ node / year
Unlimited nodes, vSphere live migration, zero-trust signal fabric.
- →Everything in Team
- →vSphere VIBs + live migration
- →Telemetry data lake (1y retention)
- →24×7 support · 1h SLA
- →Air-gapped install bundle
Sovereign
Customannual contract
Source-available licence, on-prem build pipeline, FedRAMP-aligned.
- →Everything in Datacenter
- →Source-available licence
- →Dedicated SRE pod
- →FIPS 140-3 crypto module
- →Custom kmod signing keys
Enterprise-only capabilities
Cross-node tensor parallelism over RoCEv2. Auto-rebalances when a node leaves the ring; checkpoints written to NVMe-oF every 30s.
adaptive-ctl shard --ring=8 --rdma=roce2
Lite vs Enterprise
| Capability | Lite (OSS) | Enterprise |
|---|---|---|
| Max nodes | 1 (8 cores) | Unlimited |
| QPU | Simulated, 24 qubits | IBM / IonQ / Quantinuum bridge |
| Quantisation | fp16 / int8 / int4 PTQ | + QAT, live re-quantisation |
| Networking | Loopback only | RoCEv2, SONiC, Arcos OS |
| vSphere migration | — | Signed VIB, GPU-passthrough |
| Telemetry | stdout / file | S3 data lake, 1y retention |
| Support | Community | 24×7, 1h SLA |
| Licence | Apache-2.0 | Commercial (source-available on Sovereign) |
Cluster bootstrap
bash — enterprise
1
adaptive-entlicence activate --key=$LIC --tier=datacenter
2
helminstall adaptive oci://registry.adaptive.os/charts/adaptive --values prod.yaml
3
adaptive-entqpu register --backend=ibm --token=$IBMQ_TOKEN
4
adaptive-entpolicy apply ./sla.yaml
5
# cluster online · 64 nodes · p99 72ms · governor: SLA