Coinbase fine-tuned a version of Alibaba’s Qwen3.5-9B model to detect fraud on its Onramp service. The 9-billion-parameter model outperformed frontier systems like Opus 4.5 on a proprietary benchmark of 16,140 transactions, with improvements ranging from 7.5 to 35.4 percentage points across four metrics. Median latency dropped from 1.515 seconds to 0.683 seconds, a 55% reduction, and training cost under $100 on a single NVIDIA RTX PRO 6000 GPU. The results show a 30% decrease in fraudulent transactions and a 22% reduction in fraud value, with savings estimated at three times the model’s operational cost.
Source: Read the original article
Disclaimer: this content is for information purposes only and is not financial advice. Cryptocurrencies are highly volatile: you may lose all of your capital. Always do your own research. Legal notice
