SAN FRANCISCO — Together AI has taken a transformative step in advancing generative AI capabilities with the acquisition of CodeSandbox, a pioneering platform for virtual machine (VM) sandboxes. This integration introduces the Together Code Interpreter, making the Together Inference Platform the first to feature built-in code execution capabilities. By enabling seamless code interpretation, Together AI aims to empower developers to unlock the full potential of large language models (LLMs).
CodeSandbox, launched in 2017 in Amsterdam, has earned recognition as an innovative development environment used by over 4.5 million developers monthly. Its VM sandboxes allow for instant code execution, hibernation, and resumption with superior performance, security, and scalability. With this acquisition, Together AI and CodeSandbox join forces to redefine the future of generative AI.
Revolutionizing LLM Applications
The Together Code Interpreter bridges the gap between code generation and execution. Developers can now use advanced LLMs like Llama 3.3 70B to analyze real-time data, simulate algorithms, or debug application code within an actual development environment. The technology executes LLM-generated code in secure, isolated sandboxes and feeds the results back to the LLM. This creates a powerful feedback loop, enabling advanced data analysis, visualization, and complex problem-solving.
“At Together AI, we aim to provide developers with the tools to manage the entire generative AI lifecycle efficiently and effectively,” said Vipul Ved Prakash, Founder and CEO of Together AI. “By integrating CodeSandbox’s capabilities into our platform, we’re enabling developers to build applications that execute complex logic and address real-world challenges with unmatched precision and scalability.”
Early Access to Breakthrough Technology
To extend these capabilities to a broader audience, CodeSandbox is launching the CodeSandbox SDK beta, allowing developers to create and run VM sandboxes securely at scale. This SDK offers a robust framework for integrating code interpretation into generative AI applications. Developers eager to explore this technology can access a quickstart guide to begin implementing the SDK ahead of the Together Code Interpreter’s public rollout.
“The addition of the Together Code Interpreter enhances how developers use LLMs, enabling them to solve programming problems, visualize data, and even execute agentic workflows,” said Robert Rizk, Co-Founder and COO of Blackbox AI, a company that has already integrated CodeSandbox technology into its operations. “This innovation has proven to be invaluable for handling high loads and ensuring secure, efficient code execution.”
Paving the Way for Open-Source AI
This acquisition underscores Together AI’s commitment to open-source innovation. By incorporating CodeSandbox’s technology, Together AI is enabling developers to leverage LLMs for a wider array of tasks, from drawing charts to running evaluations, all within a secure and efficient platform. The upcoming release of LlamaCoder, powered by CodeSandbox SDK and Llama 3.1 405B, exemplifies this shift, showcasing how integrated code execution enhances the development process.
“This partnership represents a new era for generative AI development,” added Prakash. “With the Together Code Interpreter and CodeSandbox SDK, we’re breaking barriers that have limited the functionality of LLMs, giving developers the tools to create solutions that are both impactful and scalable.”
By integrating secure, high-performance code execution into the Together platform, this collaboration is redefining how generative AI applications are built, deployed, and scaled. Together AI and CodeSandbox are poised to shape the future of open-source AI, ensuring that innovation remains accessible and transformative for developers worldwide.
