A startup claims it broke through a bottleneck that’s holding back LLMs

Miami-based AI startup Subquadratic has introduced a new LLM called SubQ, claiming it is faster, cheaper, and more energy-efficient than existing models. SubQ reportedly processes up to 12 times more text simultaneously, enabling it to handle data-intensive tasks like analyzing hundreds of documents or entire codebases. Despite initial skepticism, Subquadratic has released independent evaluation results from Appen, which lend credibility to its claims. Co-founder Alex Whedon noted that sharing these benchmarks earlier would have mitigated much of the doubt surrounding the model.