Sarvam reportedly trains on B300s, countering H100-based compute criticism

cneuralnetwork · x · 2026-11-21

Responding to claims that 4,000 GPUs are insufficient for from-scratch pretraining, @cneuralnetwork says Sarvam actually trains on NVIDIA B300s, arguing that assumptions based on H100s and the absence of nvpund training no longer apply to 2025-era labs.

Related event: Sarvam AI Drawn Into GPU Debate as Insider Claims B300 Training(3 posts)→

Original post →

More from Infra

Infra channel →