By 
South Africa-founded artificial intelligence startup Refiant AI has launched Protea, a new family of large language models (LLMs) capable of processing up to 10 million tokens in a single prompt, positioning the company among a growing group of AI developers pushing the boundaries of long-context language models.
The company said the new models significantly expand the volume of information that can be analysed and retained during a single interaction, enabling users to process large datasets without breaking them into smaller sections.
Protea is available in three variants with context windows of one million, five million and 10 million tokens. According to Refiant AI, the models are immediately accessible through its platform without waitlists or approval requirements.
The product launch follows the company’s $5 million seed funding round completed three months ago, led by VoLo Earth Ventures, which was intended to accelerate the expansion of its AI optimisation platform and strengthen research and development efforts.
Protea represents Refiant AI’s first major commercial product since securing the investment.
Commenting on the launch, Co-founder Mathew Haswell said the company wanted developers and businesses to gain immediate access to the technology rather than face lengthy onboarding processes.
“Customers don’t need more waitlists. They need models they can test, break and build with. Protea is live, and we want people to use it from day one,” he said.
Founded in 2025 by Mathew Haswell, Viroshan Naicker and Siddharth Gutta, Refiant AI develops machine learning technologies designed to reduce computing costs while improving the efficiency and scalability of artificial intelligence models.
According to the company, Protea’s largest model can process approximately 7.5 million words in a single prompt, opening up new possibilities for enterprise applications that require the analysis of extensive documents and complex datasets.
Potential use cases include legal professionals reviewing hundreds of contracts simultaneously, insurance companies analysing years of claims data, and software engineering teams processing entire codebases within a single AI session.
The launch comes as competition intensifies among global AI developers seeking to increase the context windows of large language models to improve reasoning, memory and enterprise performance.
Rather than focusing solely on model size, Refiant AI is positioning long-context capability as a key competitive advantage for organisations that need to analyse large volumes of structured and unstructured information efficiently.
Chief Executive Officer and Co-founder Viroshan Naicker noted that while long-context AI has attracted significant industry attention over the past year, commercially available solutions have remained relatively limited.
The company also disclosed that it has successfully demonstrated an internal prototype capable of processing up to 100 million tokens and is currently exploring pathways to benchmark and commercialise the technology.
According to Refiant AI, Protea is the first of three planned product releases as the startup continues to expand its portfolio of AI optimisation and enterprise intelligence solutions.
Comments