companies

Arcee AI

Small language model company building tools to train, route, and deploy compact enterprise LLMs as alternatives to large proprietary models.

Arcee AI was founded in 2023 in Miami by Mark McQuade, Jacob Solawetz, and Brian Benedict. McQuade and Benedict were both early employees at Hugging Face, where they worked on the company's early monetization and enterprise strategy before starting Arcee to focus specifically on small language models, models built to be cheaper and more controllable for a specific business domain rather than general-purpose giants. The company merged with the open-source model-merging project MergeKit around its seed round and has since raised roughly $29.5 million, including a $24 million Series A from investors like Emergence Capital and M12. Arcee's platform spans model training, routing, and deployment. Arcee-SuperNova, built on Meta's Llama 3.1 architecture through distillation, fine-tuning, and model merging, is positioned as a customizable alternative to closed models that enterprises can run inside their own VPC. Arcee Conductor routes prompts across Arcee's own small models and third-party models from providers like Anthropic, OpenAI, and DeepSeek based on task complexity, while Arcee Orchestra provides a no-code interface for building multi-model agentic workflows. The company has grown to roughly 35-40 employees and argues that domain-adapted small models can handle most enterprise workloads at a fraction of the cost of frontier-scale models.

Founded

2023

Headquarters

Miami, United States

Sector

lab