Migrate from ChatGPT Enterprise to Open WebUI

The cost splits across four payees: the infrastructure provider, the model provider (GPUs you run or API usage you pay for), the people who operate the stack, and any implementation partner. ChatGPT Enterprise is a per-seat subscription to OpenAI, which hosts everything. Open WebUI itself is free to run under the Open WebUI License, so the recurring spend moves to servers, inference and staff time. Ask every quote to show these lines separately so you can see which ones your own team would absorb.

One-time work includes discovery, data export, SSO and group mapping, initial configuration, testing and documentation. Recurring costs include compute and storage, GPU or API consumption, backups, patching, Open WebUI upgrades, monitoring and support hours. Open WebUI publishes frequent releases, so budget for regular upgrade testing rather than a single rollout. Ask the quote to label each line as one-time or recurring and to state the assumed release cadence.

Model choice is the largest variable. Open WebUI connects to OpenAI-compatible APIs and to Ollama, so you can pay per token for a hosted API or run open models on your own hardware. Self-hosted inference shifts cost to GPUs, power, and capacity planning; Ollama's own pricing states that running models on your own hardware is unlimited, though hardware is still your cost. For concurrency, Ollama defaults to one parallel request per model, and vLLM needs accelerator capacity sized for peak load. Ask the quote to model usage at your expected seat count and to name the inference option it assumes.

A single Open WebUI instance is simpler, but larger rollouts need more components. The scaling guide requires PostgreSQL instead of SQLite for multi-instance deployments, and multiple replicas need Redis for shared state plus shared file storage so uploads and knowledge stay reachable from every replica. Each of these adds hosting, backup and patching work. Ask whether the quote includes high availability, and which components are managed services versus self-run.

The Open WebUI License keeps the standard branding in place for deployments of up to 50 users in any rolling 30-day period. Above that threshold, altering or removing branding requires an enterprise license or written contributor permission. Enterprise licenses are quoted after contacting sales with an expected seat count, and pricing is not published. Ask your counsel to review the license against your user forecast, and ask the vendor for a written quote that states what it covers, including white-labeling and support.

Identity work depends on how many identity sources you have and how cleanly groups map to roles. Open WebUI documents SSO through OIDC and LDAP, SCIM 2.0 provisioning, and role-based access control with groups and per-resource permissions. Costs rise with custom claim mapping, multiple identity providers, and cleanup of inconsistent groups. Ask the quote to list each identity source, the group-to-role mapping it assumes, and who signs off on access reviews.

Migration effort depends on what you export and how much you need to keep. ChatGPT Enterprise admins can retrieve conversations and GPT records through the Enterprise Compliance API, and OpenAI states that deleted conversations are removed within 30 days unless legally required. Scope the work by volume of history, retention rules you must honor, audit log requirements, and whether attachments are included. Ask the quote to state the export method, the target schema, how it verifies completeness, and what is excluded.

Many teams can run Open WebUI with in-house staff, especially if they already operate containers and identity systems. The decision depends on skills, time available, and how much compliance evidence you need. DeployOpen plans, implements, integrates, hardens, backs up, upgrades, documents and supports self-hosted platforms in your own infrastructure, and most engagements are fixed-scope with agreed deliverables and timelines in writing. Ask any provider for a written scope listing deliverables, responsibilities, support hours and handover, and confirm which compliance evidence they prepare versus what your auditor issues.

The cost splits into components that are paid to different parties. Open WebUI itself is open source, so the software line is zero unless you buy an enterprise license for branding or support. The recurring costs are the servers or cloud accounts that host the interface, the database and shared storage, the model inference (either GPUs you own or per-token API charges from a model provider), and the staff time to patch and upgrade. The one-time costs are the migration work, SSO and group mapping, and testing. The biggest swing factor is inference: a per-seat subscription is replaced by hardware or metered API spend that scales with usage, not headcount. Ask any implementation partner for a quote that lists each component separately, with the assumed user count, usage volume and the hours for migration, identity integration and operations handover.

No supplied source gives a duration for this migration, so treat any single number with caution. The timeline depends on the number of workspaces and users, how much conversation history must be carried over, how many identity groups need mapping, whether models run on local GPUs or through an API, and how many environments (test, staging, production) you need. Multi-replica deployments add work, since Open WebUI's scaling guide calls for PostgreSQL, Redis and shared file storage before more than one instance runs. Ask for a written scope with deliverables and milestones before work starts, because fixed-scope engagements are agreed in writing at the outset.

Each team has a distinct role. Identity teams need to confirm how users sign in today and which provider will be the source of truth. Open WebUI documents SSO through OIDC and LDAP, plus SCIM 2.0 user and group provisioning and role-based access control with groups and per-resource permissions, so the mapping of ChatGPT Enterprise groups to Open WebUI groups must be agreed before cutover. Security teams should review API key handling, authenticator-app MFA, and how model endpoints are exposed. Platform teams own the hosting: container or Kubernetes runtime, PostgreSQL, Redis, object storage, backups and monitoring. If you plan to serve Ollama or vLLM locally, the model endpoints need network isolation, since the local Ollama API does not require authentication and vLLM's documentation recommends a reverse proxy that allowlists only user-facing endpoints.

A capable internal team can run it, provided it has people with Linux and container operations experience, identity administration skills and someone who owns the model stack. The migration itself is mostly configuration and data movement, but the skills gap usually appears in three places: getting a multi-replica deployment right, hardening model endpoints, and keeping the system patched after go-live. Open WebUI releases are frequent, with v0.12.3 published on October 16, 2026, so an upgrade process is a recurring duty, not a one-time task. Decide early whether your team will own upgrades, or whether a retainer for ongoing support and upgrades makes more sense. DeployOpen plans, implements, integrates, hardens, backs up, upgrades and documents self-hosted platforms in your own infrastructure, and most engagements are fixed-scope with the scope agreed in writing.

Workspace admins can view, export and delete end user conversations, and they can reach an audit log of conversations and GPTs through OpenAI's Enterprise Compliance API, which is the documented export path for ChatGPT Enterprise data. Plan the export as a sequence: confirm which admin roles can call the API, pull conversations and GPT records, store them in a location you control, and then import the history into Open WebUI where your format allows it. Check the limits before you rely on the export: the documented path covers what admins can retrieve, so anything outside it, such as certain attachments or workspace settings, may need a manual process. Also note that OpenAI's documented retention behavior removes deleted conversations within 30 days unless legally required, so align your export timing with any deletion schedule. Confirm the exact export fields with OpenAI before you build the import job.

Since v0.6.6, Open WebUI is distributed under the Open WebUI License, which is BSD-3-Clause style with a branding protection clause. The clause forbids removing or altering Open WebUI branding in deployments that exceed 50 users in any rolling 30-day period, unless you hold an enterprise license or written contributor permission. Code through v0.6.5 remains under BSD-3-Clause. Above 50 users, the practical question is whether you must keep the branding as delivered. If you want white-labeling or custom branding, the enterprise license covers it, and pricing is not published, so you request a quote from enterprise sales with an expected seat count. Have legal counsel read the license text and the history file before you commit, and confirm how your user count is measured.

The choice comes down to data handling, usage volume and operations capacity. Open WebUI connects to Ollama, OpenAI, Anthropic or any OpenAI-compatible provider from a single interface, so you can mix models. Using an API provider avoids GPU purchases but sends prompts to that provider, so review its data terms and your residency needs. Self-hosted models keep prompts inside your environment: Ollama runs locally and does not see your prompts or data when used that way, and it can run on macOS, Windows, Linux or Docker, while vLLM suits larger shared serving on GPUs and exposes OpenAI-compatible endpoints. Consider Ollama's documented concurrency defaults, such as one parallel request per model, before sizing a shared service. Define the model policy in the scope so you can cost each route.