Maxley vs IBM AI Assistant
A practical comparison of cost, capabilities, and infrastructure requirements for AI in IBM Maximo.
We are often asked to compare Maxley AI with the IBM AI Assistant. We think it is arguably unfair to compare these two solutions because they are so different. Most significantly, they provide dramatically different levels of AI model performance and capability to Maximo users.
However, we have provided comparative information below. Please contact us if you see any errors or have any questions. Our goal is for this information to be accurate and up-to-date.
Summary
Maxley requires zero infrastructure changes and uses leading foundation models (Anthropic Claude) selected for high agentic and inferential reasoning capabilities. Maxley further enhances Claude with a Maximo skills database, learning capability, and organizational memory. Claude has a large context window (how much can be kept in memory, reasoned over, or worked on) of 1 million tokens.
IBM AI Assistant requires new infrastructure and software platform installation and uses a compact open-weight model (gpt-oss-120b) sized to run on one or two GPUs, and answers questions through a pre-configured set of tools. This model has a much smaller context window of 131,072 tokens.
In our assessment, this means that the IBM AI Assistant and its open-source gpt-oss-120b model may be a good fit if the infrastructure build-out and cost is not an issue, and you need a high-speed agent that is well suited for routine questions, and you are okay with pre-configuring the Assistant for each data set you want to ask about.
Maxley, using Claude, will complete routine tasks as well, but also excels at complex, high-reasoning agentic work. If you want advanced analysis capability, insights offered across all Maximo applications and data, more reliable multi-step agentic processes, ad-hoc problem solving assistance, an expert-level Maximo developer (automation scripts, screens, BIRT reports, dashboards with business metrics, etc.), and reduced risk of errors and hallucinations, or if you want to avoid infrastructure costs, then you should consider Maxley.
Independent AI Benchmark test results showing Claude (Maxley) performance relative to the gpt-oss-120b (IBM AI Assistant) model.
These are general tests (they do not reflect Claude further enhanced by Maxley’s Maximo skills and knowledge database).
Maxley currently uses Claude Sonnet 5.5 for standard questions / tasks and Claude Opus 5.5 for “deep thinking” assistance.
Source: Model rankings as of October, 2026, from artificialanalysis.ai.
Comparison Details
A closer look at cost, functionality, and infrastructure requirements.
Cost
| Cost Area | IBM Maximo Assistant (9.2 + July 2026 patch) | Maxley |
|---|---|---|
| New infrastructure |
The AI Assistant requires a separate watsonx.ai OpenShift cluster
(9 nodes per IBM, including one GPU node
with two H100 GPUs recommended) plus AI
Service capacity: an estimated $15K – $22K per month on Microsoft Azure (1-year reserved to pay-as-you-go rates, Sep 2026), before storage growth, data transfer, and Cloud Pak for Data operations. Higher on AWS, where a single 2-GPU H100 is not available. [1] |
None |
| Token / Usage cost | App Points consumed to pay for tokens even when self-hosting the model | Your own Anthropic key used for tokens, no markup |
| Project to install a watsonx cluster & AI Service components | Ask your Consultant or Hosting Provider | None (Not Applicable) |
| Per-user license | None (10 App Points permanently consumed on install) | Maxley is priced per-seat, starting at a low monthly price – get your quote |
– If you use a hosting provider, a profit margin will typically be added. Ask your provider for a quote for the agentic version of AI Assistant.
– See bottom of page for reference citations.
Functionality
| Capability | IBM Maximo Assistant | Maxley |
|---|---|---|
| Answers questions about any Maximo data without pre-configuration | No | Yes |
| You can ask the AI anything | No | Yes |
| Has Maximo skills database | No | Yes |
| Has organizational memory | No | Yes |
| Has user feedback mechanism | No | Yes |
| Saved / resumable conversation history | Limited | Yes |
| Creates & hosts business metric dashboards | No | Yes |
| Language Support | English only | Any |
Support Requirements
| Requirement | IBM Maximo Assistant | Maxley |
|---|---|---|
| MAS Version Required | MAS 9.2, AI Service 9.2 with July 2026 patch (earlier 9.2 versions have higher infra costs) | Any MAS 8.x or MAS 9.x |
| Can be patched in Test before Prod? | Only with a 2nd new OpenShift cluster added for Test MAS | Yes |
| Requires new OpenShift cluster | Yes, at least one new cluster (watsonx.ai must be standalone) | No |
| Requires Cloud Pak for Data (CP4D) | Yes | No |
| Requires AI Service | Yes | No |
| Requires watsonx.ai | Yes | No |
| Requires GPU Operator | Yes | No |
References
| IBM Documentation | Reference |
|---|---|
| Deploying watsonx.ai on-premises for AI Service | View IBM Documentation ↗ |
| Deploying AI Service on-premises (cluster sizing, watsonx “own cluster” requirement, MongoDB/Db2/MinIO prerequisites) | View IBM Documentation ↗ |
| Enabling the agentic Maximo Assistant | View IBM Documentation ↗ |
See Maxley in action.
Explore real examples of how Maxley works with Maximo data, workflows, documents, development, analysis, and more.
