PANews October 12 news, Coinbase CEO Brian Armstrong said that after the company internally fine-tuned the open-weight model Qwen3.5-9B using proprietary data, it outperformed frontier large models on the specialized task of fraud prevention, while also being lower in cost and faster. Armstrong said that this model will become more common within enterprises in the future: frontier models excel at general tasks, while specialized models for niche domains will perform better. Coinbase is simplifying the process for internal employees to fine-tune models, even providing internal skill instructions to help quickly start training, and expects that some backend services can be completely replaced by this type of architecture.
Coinbase CEO: After fine-tuning the Qwen3.5-9B model, fraud prevention delivers lower cost and faster speed.
Noticias relacionadas
Coinbase obtuvo un beneficio neto de 700 millones de dólares en el tercer trimestre, con casi el 90% del beneficio procedente del aumento de los precios de las criptomonedas.
Cuando el precio de Bitcoin está atascado en un rango, el mercado adoptará una visión más sobria de la rentabilidad y la resiliencia al riesgo de Coinbase como...