NeuroDX Unveils 240M Model Sphota Outperforming Liquid AI On Benchmarks

2 Nov 2026, 12:58 pm · 3h ago · 1 min read · Analytics India Magazine

Indian AI startup NeuroDX, backed by the IndiaAI Mission, has developed Sphota, a 240-million-parameter small language model that reportedly outperforms Liquid AI across 11 benchmark tests. According to CEO Siddharth Panwar, Sphota achieved an average score of 52.4% compared to Liquid AI’s 52.1%, while utilizing 50 times less training data and delivering 4.5 times greater memory efficiency. Originating from research on brain electroencephalogram (EEG) signals, the model’s novel architecture focuses on capturing semantic meaning directly instead of predicting next tokens, providing an alternative, resource-efficient pathway for domestic foundation AI model development.