Access any model, anywhere on watsonx.ai

Written by Nisarg Patel , Maryam Ashoori, PhD , and AJ Albanese, MBA

We’re excited to announce the upcoming watsonx.ai Model Gateway in Public Preview, a new enterprise solution that streamlines access to leading AI foundation models through a unified interface.

With Model Gateway, organizations can integrate IBM’s Granite models alongside industry-leading foundation models from OpenAI, Anthropic, NVIDIA, Cerebras and more without vendor lock-in. This AI-agnostic approach enables businesses to maintain control, enhance flexibility and optimize costs, regardless of where models are hosted.

Watsonx.ai’s multi-model strategy

Model Gateway is the final pillar in watsonx.ai’s multi-model strategy, giving customers the ultimate choice in using third-party hosted models on platform. Users can leverage third-party models to develop agents using any OOTB templates, and quickly deploy them as AI services.

watsonx.ai Model Gateway enables:

• Model Choice: Access to 1000+ SOTA models, whether hosted on watsonx.ai or in another third-party environment
• Opt-In Mechanisms: Models access via the Gateway must be provisioned, so all connections are opt-in to maintain the proper access
• Deployment Flexibility: Model Gateway will be available on-Premises once generally available, allowing enterprises to have connection to non-hosted models

Managing the complexity of multiple foundation models

As businesses scale AI adoption, many leverage multiple foundation models across cloud, on-premises and hybrid environments. Managing this complexity presents challenges in governance, cost-efficiency and operational oversight. Enterprises need a centralized AI control plane that simplifies model orchestration, enforces policies, optimizes costs and provides near real-time visibility.

Over the coming months, developers will be able to leverage Developer Playground tools like Prompt Lab to experiment with any foundation model and deploy AI applications efficiently, develop RAG solutions using AutoAI RAG, and integrations with additional products in the watsonx family. Additional features will be enabled in the coming weeks as well to enforce more controls for the developer.

Seamlessly switch between models

With a uniform API approach using an OpenAI compatible API, businesses can seamlessly switch between models without the need for costly reconfiguration.

As part of our mission to bring our users with more options for foundation model serving, watsonx.ai Model Gateway also connects to Cerebras, a leading provider of fast inference using their own wafer-scale chips.

"We are excited to bring Cerebras' record-setting AI inference capabilities to IBM’s watsonx.ai Model Gateway,” said Alan Chhabra, EVP, Worldwide Partners, of Cerebras.”This collaboration directly addresses the growing demand for high-performance AI, allowing enterprises to seamlessly deploy powerful foundation models for deep reasoning and agentic AI without latency bottlenecks. It's a game-changer for businesses seeking to operationalize their most ambitious AI initiatives at scale."

We’re excited to share this feature with you and are looking forward to receiving your feedback to help us iterate.

Try watsonx.ai for free today

Explore the end-to-end tutorial

Read our documentation

Statements regarding IBM's future direction and intent are subject to change or withdrawal without notice and represent goals and objectives only.