IBM Bob’s self-hosted option runs its backend on an enterprise’s Red Hat OpenShift cluster, while developers continue working in Bob IDE and Bob Shell. Teams can connect that backend to a supported external model service using their own account, or run supported open-weight models on customer GPUs—including in a fully disconnected environment. Those choices change where code context is processed and how much infrastructure and security work the organization must operate.
What IBM Bob self-hosted means
Self-hosted does not mean installing a standalone Bob desktop app on each developer’s machine. The Bob backend runs as a workload on the organization’s own OpenShift cluster. Developers retain the Bob IDE and Bob Shell experience, while the organization operates the underlying infrastructure, services, integrations, and lifecycle.
As an Amazon Associate I earn from qualifying purchases.
Free tools Windows power users keep installed
One-click scans. No signup required.
IBM says the backend package includes identity, an inference gateway, audit logging, and usage metering, managed by an operator. That package is not a substitute for a complete enterprise security monitoring system: IBM’s documentation says security event logging and monitoring are handled at the OpenShift platform level, not provided by Bob. See IBM’s self-hosted overview documentation for the deployment model and platform responsibilities.
Choose where model inference runs
The two deployment routes are not equivalent for data handling. In both, the Bob backend remains on the customer’s OpenShift cluster; the difference is where inference runs, how requests travel, and who controls access to the model.
#1 Best Overall
| Route | Where inference runs | Network path and model access | Operational implication |
|---|---|---|---|
| Hybrid or private SaaS connectivity | At an external model service. | Bob sends model requests, including the code context they carry, to a service accessed through the organization’s own account. IBM’s September 30 announcement lists Claude Sonnet 5.0, Claude Opus 4.8, Gemini 3.7 Flash, and OpenAI GPT 5.6 Sol as supported choices; models are customer-provided, and IBM describes an eligible bring-your-own-license approach. | The organization operates the Bob backend and controls its model-service account. The external request path must be assessed against internal data-handling rules. |
| Fully self-hosted models | On customer GPUs. | Supported open-weight models run on customer infrastructure; IBM names NVIDIA Nemotron 3 Ultra and Poolside Laguna S 2.1. This route can serve a network with no outbound connection. | The organization operates both the Bob backend and the model-serving hardware, including capacity planning. |
IBM says it plans to expand the supported model portfolio and introduce multi-model routing in future releases. Those are plans, not capabilities to assume are available today. Confirm the current model list, licensing eligibility, and configuration with IBM.
How to approach a self-hosted deployment
- Decide the data and connectivity boundary. Determine whether policy allows code context to leave the organization for an external model service, or whether inference must run on customer GPUs inside a disconnected environment.
- Confirm the platform and model combination. Identify the OpenShift environment, intended model, context length, and number of concurrent developers. IBM says GPU sizing depends on model-vendor guidance, quantization, context length, and concurrency; the cited material does not establish a universal server configuration.
- Plan installation and image transfer. IBM’s installation CLI can mirror images into a private registry and transfer images from a connected system into an isolated environment. IBM describes a single-node OpenShift installation as sufficient for a proof of concept, not as production sizing guidance.
- Validate production requirements and controls. Confirm supported configurations and production system requirements against IBM’s current installation documentation for the chosen workload. Map identity, network, storage, lifecycle, security event collection, and monitoring responsibilities to the teams operating OpenShift.
- Arrange access and commercial terms. IBM’s release directs organizations to an IBM representative or Business Partner for a demo or proof of concept. Ask IBM to confirm entitlements, any premium packages, support terms, and pricing; the cited announcements do not state an exact self-hosted price.
Self-hosted versus IBM Bob SaaS
The central trade-off is operational control versus operational responsibility. IBM describes SaaS as hosted and managed by IBM, with automatic updates. Self-hosted places the backend on the customer’s OpenShift cluster and shifts lifecycle operations and security configuration to the customer.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
| Consideration | IBM Bob SaaS | IBM Bob self-hosted |
|---|---|---|
| Backend infrastructure | Hosted and managed by IBM. | Runs on the customer’s OpenShift cluster. |
| Updates and lifecycle | IBM manages the service and automatic updates. | The customer manages lifecycle operations. |
| Security configuration | Managed as part of IBM’s hosted service; confirm the controls relevant to your requirements with IBM. | The customer configures and operates security at the platform level, including security event logging and monitoring. |
| Data path | Depends on the service and model arrangement; verify the applicable data handling with IBM. | Depends on the chosen model route: external service requests carry code context, while customer-hosted inference can remain in a disconnected network. |
Self-hosting can suit teams with strict infrastructure or data-boundary requirements, but it does not remove operational work. It relocates that work to the organization’s platform and security teams.
Optional modernization packages
IBM identifies optional packages for Java Modernization, IBM i, and IBM Z. The self-hosted IBM Z package is intended for Z application understanding and modernization, with Z-specific context and analysis. Confirm package availability and commercial terms for the deployment you are planning.
Rank #3
- Used Book in Good Condition
Availability and enterprise context
IBM’s release post says general availability began September 24, 2026; its news announcement was dated October 1, 2026. The route to access is enterprise-oriented: IBM points prospective customers to its representatives or Business Partners for demos and proofs of concept.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Scan for outdated or missing drivers - takes under a minute3Repair Windows errors before they cause bigger problemsIn its October 1 announcement, IBM attributed two figures to outside research: Futurum Research projected that hybrid and edge deployments would account for 44% of the AI infrastructure market by 2030, versus 46% for public cloud; and the IBM Institute for Business Value reported that 68% of surveyed executives found it challenging to meet data residency and sovereignty requirements across geographies in its June 2026 report, The Calculus of AI Sovereignty. These figures provide context for interest in deployment control; they are projections and survey responses, not proof that a particular Bob deployment meets a specific compliance requirement.
Quick Recap
Best Value
- IBM X3550 M4 4B Server
- 2x 2.50GHz E5-2640 12-Cores Total
- 32GB RAM / No Hard Drives / No Hard Drive Trays
- M5110 w/ 1GB
- No Operating System
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
