Peter Zhang
May 05, 2025 22:08
Arcee AI migrates from AWS to Together Dedicated Endpoints, optimizing costs and performance for their specialized small language models, enhancing operational agility and efficiency.
Arcee AI, a company focused on simplifying AI adoption, has made a strategic move by transitioning its specialized small language models (SLMs) from Amazon Web Services (AWS) to Together Dedicated Endpoints. This migration, according to together.ai, has brought significant improvements in cost efficiency, performance, and operational agility for Arcee AI.
Optimizing Small Language Models
At the heart of Arcee AI’s strategy is the development of specialized small language models optimized for specific tasks, typically under 72 billion parameters. The company leverages proprietary techniques for model training, merging, and distillation to produce high-performing models that excel in tasks like coding, text generation, and high-speed inference.
With the migration to Together AI, seven of these models are now accessible via Together AI’s serverless endpoints. These models include Arcee AI Virtuoso-Large, Arcee AI Virtuoso-Medium, Arcee AI Maestro, Arcee AI Coder-Large, Arcee AI Caller, Arcee AI Spotlight, and Arcee AI Blitz, each designed for various complex tasks ranging from coding to visual tasks.
Software Enhancements: Arcee Conductor & Arcee Orchestra
Additionally, Arcee AI has developed two software products, Arcee Conductor and Arcee Orchestra, to enhance their AI offerings. Conductor serves as an intelligent inference routing system, efficiently directing queries to the most suitable model based on task requirements. This system not only reduces costs but also improves performance benchmarks by utilizing the best model for each task.
Arcee Orchestra focuses on building agentic workflows, enabling enterprises to automate tasks through seamless integration with third-party services. The no-code interface allows users to create automated workflows effortlessly, powered by AI-driven capabilities.
Challenges with AWS and the Move to Together AI
Initially, Arcee AI deployed its models via AWS’s managed Kubernetes service, EKS. However, this setup posed challenges, requiring significant engineering resources and expertise, making it cumbersome and costly. AWS’s GPU pricing and procurement difficulties further complicated matters, prompting Arcee AI to seek alternative solutions.
Together Dedicated Endpoints offered a managed GPU deployment, eliminating the need for in-house infrastructure management. This transition simplified Arcee AI’s operations, providing greater flexibility and cost-effectiveness. The migration process was seamless, with Together AI managing the infrastructure and providing API access to Arcee AI’s models.
Performance Gains and Future Prospects
Post-migration, Arcee AI reported performance improvements across its models, achieving over 41 queries per second and reducing latency significantly. These enhancements have positioned Arcee AI to continue expanding its offerings and innovating within the AI landscape.
Looking ahead, Arcee AI plans to further integrate its models with Arcee Orchestra and enhance Arcee Conductor with specialized modes for tool-calling and coding. Together AI remains committed to optimizing its infrastructure to support Arcee AI’s growth, ensuring superior performance and cost-efficiency.
This partnership reflects the evolving dynamics of the AI industry, where companies like Arcee AI leverage cloud-based solutions to refine their offerings and deliver better return on investment. For more details, visit together.ai.
Image source: Shutterstock
Credit: Source link
