Zemo is an enterprise-focused AI model developed by Zetaris, built on the gpt-oss-120b foundational model and adapted for enterprise data, analytics, and AI engineering use cases.
Zemo is designed to understand business-oriented data requests and assist with tasks such as natural-language-to-SQL, enterprise data analysis, semantic data interpretation, and AI agent workflows.
Zemo can be deployed in controlled enterprise environments using high-performance inference engines such as vLLM and SGLang.
Zemo combines the capabilities of the underlying gpt-oss-120b foundation with Zetaris-specific optimisation and enterprise data intelligence capabilities.
The model is designed to support:
Zemo is intended to operate as part of the broader Zetaris Data & AI Platform, where AI models can interact with governed enterprise data through semantic layers, data virtualisation, tools, and AI agents.
Zemo is built on:
Foundation Model: gpt-oss-120b
The underlying foundation model provides the general-purpose language and reasoning capabilities, while Zemo is designed and configured for enterprise data and AI workloads.
Zemo is not intended to replace the foundation model. It extends the foundation model for specific enterprise data and AI engineering scenarios.
Zemo can translate business questions into structured query logic.
For example:
"Give me a table by discharge country showing where product has been sold year to date 2026."
Zemo can interpret the request in terms of:
and generate the appropriate query against the underlying data model.
Zemo can convert natural-language business rules into executable query logic.
For example:
"Calculate the invoice value in USD. If the invoice currency is USD, use the invoice value; otherwise set the USD value to zero."
This can be represented as:
CASE
WHEN CurrCode = 'USD'
THEN CAST(InvItemVal AS DECIMAL(15,2))
ELSE 0
END AS invoice_value_usd
Zemo is designed to work with enterprise concepts such as:
The model can use metadata and semantic definitions to bridge the gap between business language and underlying data structures.
Zemo can be deployed using modern high-performance inference engines.
Zemo can be hosted using vLLM for scalable model serving and OpenAI-compatible inference APIs.
Example:
vllm serve <zemo-model>
Once deployed, applications can interact with the model through the inference API.
Zemo can also be deployed using SGLang, providing an alternative high-performance serving architecture.
python -m sglang.launch_server \
--model-path <zemo-model>
The choice between vLLM and SGLang can depend on the target workload, hardware configuration, batching requirements, latency requirements, and serving architecture.
Zemo is designed to operate as the intelligence layer within an enterprise data architecture.
Business User
|
v
Natural Language Request
|
v
+----------------+
| Zemo |
| Enterprise AI |
+----------------+
|
+---------------+---------------+
| |
v v
Semantic Layer AI Agents / Tools
| |
+---------------+---------------+
|
v
Zetaris Data Platform
|
+----------------+----------------+
| | |
v v v
Databases APIs Data Lakes
This architecture separates AI reasoning from enterprise data access, allowing data access and governance to remain under the control of the enterprise data platform.
Zemo is designed for organisations that require control over where their AI models run and how enterprise data is accessed.
The model can be deployed within controlled infrastructure, including private cloud, on-premises, or enterprise Kubernetes environments, subject to the supported deployment configuration.
This enables organisations to build AI applications without necessarily sending enterprise data to an externally hosted model endpoint.
Zemo is intended for:
Prompt:
Show product sales by discharge country for year to date 2026.
Potential output:
Discharge Country | Sales | Quantity
------------------|-------|---------
Australia | ... | ...
Singapore | ... | ...
India | ... | ...
United States | ... | ...
Prompt:
Calculate invoice value in USD. Only include invoices where the currency is USD.
Zemo can generate the corresponding business logic and SQL.
Prompt:
Create a derived field that calculates the net invoice value after discounts and excludes VAT.
Zemo can translate the business requirement into the appropriate query logic based on the available schema and semantic definitions.
Zemo can serve as the reasoning component within an AI agent architecture where agents can:
| Attribute | Details |
|---|---|
| Model | Zemo |
| Organisation | Zetaris |
| Foundation Model | gpt-oss-120b |
| Model Type | Enterprise AI / Data Intelligence |
| Primary Domain | Enterprise Data & AI |
| Inference | vLLM / SGLang |
| Deployment | Private / Controlled Enterprise Environments |
| Key Applications | SQL, Analytics, Data Engineering, AI Agents |
Zemo can generate incorrect or incomplete responses. Generated SQL, calculations, and analytical conclusions should be validated before being used in production or for critical business decisions.
Performance may depend on:
Zemo should therefore be deployed with appropriate validation, monitoring, access control, and human oversight.
Zetaris recommends that organisations deploying Zemo:
Zetaris is an enterprise data and AI platform providing unified access to data across databases, APIs, data warehouses, data lakes, and other enterprise systems.
Zetaris combines data virtualisation, semantic data access, integration, analytics, and AI to help organisations build AI applications and agents over governed enterprise data.
Learn more about Zetaris:
Zemo is built on the gpt-oss-120b foundational model.
Please refer to the associated foundation model documentation and licence for applicable attribution and usage requirements.
The use, distribution, and modification of Zemo are subject to the licence specified for this model repository.
Please review the licence before using Zemo in commercial or production environments.
Zemo — Enterprise AI by Zetaris
Built for Data. Designed for Enterprise AI.