Plans change as we learn from customers, so items below carry no dates and are not commitments until they ship. Shipped changes are in the release notes. Tell us what matters to you through support.

Available today

  • OpenAI-compatible chat (streaming, tools, JSON output, image input), embeddings with batch jobs, image generation and editing, text-to-speech, transcription with subtitles, live captions, voice cloning and computer vision.
  • API keys with scopes and separate administration rights; HTTPS; governance (rate limits, quotas, budgets, scheduling, content rules, moderation, model lifecycle, region policy); signed audit export.
  • AI Gateway: five routing strategies, model-aware routing, failover, canary rollouts, DNS and local-network discovery, backpressure and zero-downtime upgrades.
  • Windows app with a Windows service; Docker images; Helm chart and Kustomize manifests; bring-your-own-licence offers on the Azure and AWS marketplaces.
  • Built-in dashboard, Prometheus metrics, usage reports and redacted logs.

Next

ItemWhy
/v1/responses, /v1/completions and /v1/moderationsNewer OpenAI clients and older tools use them.
Image, speech and vision models listed in /v1/models with their capabilitiesClients discover every model type from one call.
Compressed audio (MP3, M4A) accepted by local transcriptionToday local models take WAV only.
A published OpenAPI description of the APICode generation and API gateways.
Downloading a model without making it activeSafer model installs on a busy shared server.
A tamper-evident chain across audit recordsDetect a removed or edited record, not only an altered export.
macOS and Linux desktop appsToday Linux runs the containers, and there is no macOS server app.

Being considered

  • Longer licence terms for sites with very restricted internet access (Enterprise).
  • Routing by data-residency zone and by organisation across a gateway farm.
  • A fleet view of many servers in one app.
  • More cloud marketplaces.

Announced changes

ChangeDateWhat to do
The legacy local-AI API on port 11434 stops2026-12-31Move integrations to /v1 — see migrate.
AI Gateway requires Pro Commercial; gateway nodes use no seat2026-10Nothing, if you run a gateway on Commercial.
New API keys no longer get administration rights by default2026-10Issue monitoring tools a key with Allow server administration.

What we do not plan to do

  • Charge per request or per token. AI Server is priced per server.
  • Route your data through our cloud. The server's data path stays between your clients and your server.