Collaboration between AI21 and Together AI

Enterprise adoption of open-source AI is still moving slowly. McKinsey found that more than 75% of organizations plan to expand their use of open-source AI, while Menlo Ventures' 2025 mid-year LLM market update tallied that only 13% of AI workloads actually run on open-source models. AI21 Labs and Together AI announced a collaboration to narrow this gap.

The integration combines AI21 Maestro with Together AI's high-performance open-source model platform, simplifying how enterprises evaluate, route, and operate models. In particular, it provides native access to the Jamba open-source family, letting teams work with multiple models in one environment without rewriting existing workflows.

Within Maestro, teams can view capability summaries for each model and assign the right model to each step of an agentic workflow. As needed, they can use Jamba models or switch to Together AI's open-source LLMs, and it's also possible to combine multiple models.

• Together AI: provides an optimized open-source model ecosystem.
• AI21 Maestro: handles orchestration, governance, and stable execution.
• Jamba: an enterprise-grade open-source family supporting long-context reasoning.

Both companies emphasize that this collaboration delivers both the openness of open source and the control and reliability enterprises need for operations. The key lies in making model choice truly operational in production environments.