South African startup Refiant unveils Protea AI
South African startup Refiant has launched a new family of large language models that can process up to 10 million tokens in a single prompt. The release comes as businesses look for AI tools that can handle larger datasets. The models, called Protea, are available with context windows of one million, five million and 10
South African startup Refiant unveils Protea AI
South African startup Refiant has launched a new family of large language models that can process up to 10 million tokens in a single prompt. The release comes as businesses look for AI tools that can handle larger datasets. The models, called Protea, are available with context windows of one million, five million and 10 million tokens. A larger context window allows a model to retain more information during a session. This reduces the need to split long documents or large codebases into multiple prompts.
Backed by fresh funding
The launch follows Refiant’s $5 million seed funding round announced three months ago. The investment, led by VoLo Earth Ventures, is supporting the startup’s work on improving how large language models use computing resources. Founded in 2025 by Mathew Haswell, Viroshan Naicker and Siddharth Gutta, Refiant develops software that helps AI models process more information while using less computing power.
Built for larger workloads
According to the startup, the largest version of Protea can hold the equivalent of about 7.5 million words in a single prompt. That could help legal teams review large collections of contracts. Insurers could analyse years of claims records, while software developers could work through entire codebases without repeatedly uploading files. Refiant has made the models available through its platform from launch. Users do not need to join a waiting list or request approval. “Customers don’t need more waitlists,” said co-founder Mathew Haswell. “They need models they can test, break and build with. Protea is live, and we want people to use it from day one.”
A growing focus on longer context
The launch comes as AI companies compete to increase the amount of information their models can process in a single request. Larger context windows are becoming more important for businesses working with lengthy documents, technical records and large datasets. Naicker said long context models have received plenty of attention over the past year. However, few have been widely available for commercial use. Refiant says it has also built an internal prototype capable of processing up to 100 million tokens. The company has not said when that technology could be be released. Protea is the first of three products it plans to launch.
