New AI Model Breaks Ground with Massive Context Window

Refiant AI, a South Africa-based startup specializing in algorithmically compressing AI models, has unveiled Protea, a suite of large language models (LLMs) capable of processing up to 10 million tokens in a single prompt. This represents a significant leap forward in AI memory capacity compared to existing solutions.

The Protea models come in three versions offering context windows of one million, five million, and the flagship 10 million token capacity. Notably, access is granted immediately through Refiant’s platform without requiring waitlists or approvals—a deliberate move by the founders to encourage widespread testing and adoption.

“Customers don’t need more waitlists,” said Mathew Haswell, co-founder of Refiant AI. “They need models they can test, break, and build with. Protea is live, and we want people to use it from day one.”

The launch follows a $5 million seed funding round led by VoLo Earth Ventures, which the company will use to expand its platform and deepen research partnerships. ## What This Means for Users

At full capacity, Protea can retain roughly 7.5 million words—equivalent to analyzing entire books or extensive datasets in one go. Refiant AI highlights several practical applications:

  • Legal teams: Reviewing hundreds of contracts simultaneously
  • Insurers: Analyzing years of claims data for patterns and insights
  • Engineering teams: Processing complete software codebases for optimization
  • Financial analysts: Evaluating complex portfolios with full historical context

The company envisions these capabilities empowering enterprises to make more informed decisions based on comprehensive analysis. ## Contextualizing the Innovation

While other AI providers are also expanding context windows, Protea’s 10 million token capacity sets a new benchmark. Anthropic’s Claude offers up to 500,000 tokens in certain enterprise tiers, while Google’s Gemini provides one million tokens on higher-tier plans.

Refiant AI is positioning this expanded memory as a competitive advantage for businesses dealing with large datasets—allowing them to uncover deeper insights than previously possible.