The company published the roundup on October 21, 2026, alongside GTC Berlin.
According to NVIDIA, Revolut is rebuilding its stack around its PRAGMA foundation model, initially serving 80 million customers and engineered to scale threefold. Starling Bank trained an agentic model on Civo's UK sovereign cloud using its transaction data, while Lloyds Banking Group built its ATUM transaction foundation model on RTX PRO 6000 Blackwell GPUs. NVIDIA also said bunq's Finn assistant uses open-weight models at 60% of the cost per answer of leading closed models.
The post names further projects at Santander, NatWest, Modulr, LSEG, HSBC, and Stripe. It says Stripe recorded 42x faster data pipelines and 92% cost savings for offline testing of its Radar fraud-detection pipelines.
These are vendor-reported deployment descriptions and performance figures. The source does not provide independent validation, evaluation methods, deployment volumes beyond Revolut's stated customer base, model weights, or evidence that reported offline results translate to production outcomes.