Refiant, an AI optimization firm with South African origins, has launched its Protea suite of long-context AI models. This new offering features a context window up to ten times larger than leading tools like ChatGPT and Claude. The launch marks a significant step in AI's ability to process and remember vast amounts of information at once.
A New Benchmark for AI Context
A large language model's context window dictates the amount of text it can process simultaneously. While top models offer windows up to one million tokens, Protea expands this capacity to an unprecedented 10 million. This leap allows the model to handle longer inputs and generate more contextually aware outputs.
This massive context window is equivalent to processing 7.5 million words or 15,000 pages in a single instance. Protea can analyze five years of emails or 83 novels in one conversation without losing any information. This capability removes previous constraints that forced developers to fragment large datasets for AI analysis.
Unlocking Enterprise Potential
Refiant's Protea is poised to unlock previously impossible enterprise use cases by handling complex information at scale. For example, engineering teams can feed an entire codebase into the model, condensing a month of analysis into a single day. This dramatically accelerates development cycles and improves the quality of software analysis and debugging.
The applications extend across various industries, from finance to law. Insurance firms can upload years of claims data to identify subtle fraud patterns, while legal teams can analyze vast case files for relevant precedents. The model's ability to maintain full fidelity over large datasets is a game-changer for data-intensive professions.
Furthermore, Protea effectively solves the "lost in the middle" problem, a common limitation where models lose accuracy on information buried within long texts. This ensures consistent and reliable performance, which is crucial for building robust agentic workflows. These agents can now operate across extensive contexts without losing track of earlier reasoning.
A Nature-Inspired Approach
Refiant was founded in 2025 by Dr. Viroshan Naicker, Siddharth Gutta, and Mathew Haswell, who believed modern AI was inefficient. Their methodology is inspired by nature, utilizing evolutionary search and swarm-style optimization to find efficient solutions. This approach prioritizes clever optimization over brute-force computational power, a philosophy reflected in Protea's design.
The company's innovative techniques first gained attention when they compressed a large OpenAI model to run on a standard laptop. This success helped secure a $5 million seed round in April 2026, led by VoLo Earth Ventures. Refiant also established research partnerships with Imperial College London and University College London.
Immediate Availability and Future Vision
Unlike competitors who have only promised similar capabilities, Refiant has made Protea immediately available to the public. Users can access models with one, five, and 10 million token windows today without a waitlist or approval process. Co-founder Mathew Haswell stated that customers need models they can test and build with from day one.
This beta launch represents the first step in a three-stage roadmap, with more announcements expected over the next three months. The platform is currently free, with plans to introduce commercial tiers in the future. The team is encouraging users to stress-test the system across various industries to explore its full potential.
The launch of Refiant's Protea represents a pivotal moment for artificial intelligence, significantly expanding the boundaries of what models can comprehend. By providing immediate, open access to a 10 million-token context window, the company is empowering developers to tackle previously insurmountable data challenges. This development showcases a more efficient approach to AI and sets a new standard for the industry.