Skip to content
Apex
System status

All systems operational.

Live · 2026-09-29

We publish live status for every service we operate, plus a rolling history of changes. For paid customers, the same signals are available as Prometheus and OpenTelemetry streams into your environment.

Inference

  • Aether API · 7BOperational
  • Aether API · 40BOperational
  • Aether API · 280B sparse-MoEOperational
  • Aether Bio · 12BOperational
  • Aether Edge · 1.3B (Preview)Operational

Platform

  • Agent runtimeOperational
  • Memory and trace storeOperational
  • Engineering disciplineOperational
  • Semiconductor disciplineOperational
  • Drug discovery disciplineOperational
  • Autonomous-lab orchestratorAcoustic-dispenser driver upgrade rolling out — invisible to callersOperational

Surfaces

  • Software-engineering adaptersOperational
  • Customer consoleOperational

Site

  • Documentation siteOperational
  • Marketing siteOperational
Service-level objectives

The numbers we are on the hook for.

Targets are public. Trailing 30-day actuals are computed continuously. Error budget is the customer's lever — when we burn through it, we slow down.

ServiceTargetTrailing 30dBudget used
Aether API · availability99.95%99.98%12%
Aether API · p50 first-token latency≤ 280ms210ms—
Aether API · p99 first-token latency≤ 2.0s1.7s—
Agent runtime · job durability99.999%99.999%3%
Memory store · read availability99.99%99.995%8%
Customer console · availability99.9%99.94%21%
Scheduled maintenance

What's coming up.

We publish every planned maintenance window with at least 72 hours of notice. Customers under SLA also receive direct notification through their account contacts.

02:00 – 02:30 UTC

Memory and trace store

Schema migration to support per-tenant retention policies. Read availability maintained; writes briefly fall back to local buffer.

01:00 – 03:00 UTC

Aether API · 280B sparse-MoE

Promotion of Aether 1.0.4 with the long-rollout fix. Blue/green deploy; eval gate must pass for promotion.

Recent events

The last few events.

We publish a short note for every change that touched a customer surface — and a longer postmortem for anything that broke one.

  1. Sev-4

    Acoustic-dispenser driver upgrade

    Autonomous-lab orchestrator · 22 min

    Rolled out an upgraded acoustic-dispenser driver in three regions. No customer-visible impact; one queued plate held briefly during the swap.

  2. Sev-3

    Aether 1.0 promotion

    Aether API · all sizes · 36 hours (planned)

    Promoted Aether 1.0 to GA. Inference traffic shifted behind percentage rollout; rollback path retained for a week. No customer-visible regressions detected during ramp.

  3. Sev-2

    Memory-store secondary failover

    Memory and trace store · read availability · 9 min

    A regional failover in our cloud provider's storage layer briefly disrupted read availability in EU-West. Reads served from secondary; no writes lost. Postmortem published.

    Read the postmortem
  4. Sev-4

    Drug-discovery surface blue/green deploy

    Customer console · 2 min

    Migrated control-plane to the new blue/green deploy pipeline. Two minutes of read-only API during cut-over, scheduled and announced.

  5. Sev-2

    Drug-discovery agent regression

    Drug discovery discipline · 1h 18m

    A regression in the docking agent caused inflated docking scores for a subset of structures. Detected by the eval-gated promotion guard, rolled back automatically. Affected customer runs were re-executed at no charge.