IBM watsonx Orchestrate now supports GPT-OSS 120B on Amazon Bedrock

As AI agent deployments grow across enterprise systems, organizations need a way to bring together foundation models, tools, and governance frameworks to scale AI reliably across business functions.

IBM watsonx Orchestrate addresses this by bringing AI agents, business workflows, and enterprise integrations together in a single governed environment. It is available as a fully managed SaaS through AWS Marketplace, including a free trial.

Today, IBM watsonx Orchestrate expands its foundation model options with the addition of OpenAI GPT-OSS 120B, available through Amazon Bedrock. This gives customers more flexibility in how they build and configure AI agents without managing additional model infrastructure.

What is IBM watsonx Orchestrate?

Building effective AI agents requires more than choosing the right model. Organizations need agents that connect to the right systems, follow governance policies, and scale across business functions without constant oversight.

IBM watsonx Orchestrate is designed for that. It gives organizations a single, governed environment to build, deploy, and manage AI agents across business functions, from HR and sales to procurement and customer care.

IBM watsonx Orchestrate provides:

• Unified multi-agent orchestration. Coordinate AI agents built natively in watsonx Orchestrate or imported from external frameworks or agentic platform like Amazon Bedrock AgentCore, giving you flexibility in how you design and connect your agent workflows.
• Flexible agent development for every user:
  • Business users: Use Agent Builder, a no-code studio, to create custom agents, customize more than 100 domain-specific agent templates, and integrate knowledge sources and workflows.
  • Developers: Use the Agent Development Kit (ADK) to build agents and tools using Python and OpenAPI, add custom integrations, and deploy across watsonx Orchestrate offerings.
• Conversational AI interface. Interact with agents using natural language through chat, voice, or integrated channels such as Slack and Microsoft Teams, making it straightforward for any team member to delegate tasks to agents.
• Enterprise AI Agent governance. Observe and manage agents centrally with guardrails, policy enforcements and role-based security access controls , so organizations can deploy AI agents confidently and at scale.

Regional availability

IBM watsonx Orchestrate SaaS is available in the following AWS Regions as of March 2026:

What is GPT-OSS 120B?

GPT-OSS 120B is an open-weight model from OpenAI built for the kind of complex, multi-step work that enterprise AI agents demand.

Despite having 117 billion total parameters, GPT-OSS 120B uses a Mixture of Experts (MoE) architecture that activates only a fraction of those parameters for each request. In practice, this means you get the reasoning capability of a large model with the inference efficiency needed to run agents at enterprise scale — without the cost or latency typically associated with models of this size.

For agentic use cases, GPT-OSS 120B brings several capabilities that matter:

• Long-context understanding. A 128,000-token context window lets agents process lengthy documents, conversation histories, and multi-step instructions in a single pass.
• Transparent reasoning. Chain-of-thought output gives you visibility into how the model reaches its conclusions, which is valuable when you need to audit or validate agent decisions.
• Reliable tool use. Native function calling and structured output support make the model well suited for agents that coordinate across multiple enterprise systems to complete business tasks.

Together, these capabilities make GPT-OSS 120B a strong fit for the multi-step, tool-driven workflows that IBM watsonx Orchestrate is designed to orchestrate.

Amazon Bedrock as a model provider in IBM watsonx Orchestrate

IBM watsonx Orchestrate provides access to foundation models directly as part of the SaaS. You do not need a separate Amazon Bedrock account or subscription to use GPT-OSS 120B. IBM manages model access through Amazon Bedrock on your behalf, so you can focus on building agents rather than managing model infrastructure. When you configure an agent in the Agent Builder, you can now select GPT-OSS 120B with Amazon Bedrock as the provider.

The following screenshot (figure 1) shows the IBM watsonx Orchestrate Agent Builder with GPT-OSS 120B and Amazon Bedrock configured as the provider.