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

SubQ reportedly can process up to 12 times more text simultaneously, enabling it to handle data-intensive tasks like analyzing extensive document sets or entire codebases. Subquadratic asserts that SubQ achieves performance comparable to leading models from Google DeepMind, OpenAI, and Anthropic on critical tasks such as coding.

Initially, the company faced skepticism due to a lack of extensive evidence beyond self-published test scores and limited public access to the model. However, Subquadratic has since released results from independent evaluations conducted by third-party firm Appen, aiming to validate its claims.

Subquadratic cofounder and chief technology officer Alex Whedon acknowledged the initial skepticism, stating that releasing third-party benchmarks earlier would have been beneficial.