[Coinbase CEO: Fine-Tuned Open-Source Small Model Outperforms Cutting-Edge Models in Anti-Fraud Tasks] Coinbase CEO Brian Armstrong posted on the X platform stating that Coinbase has fine-tuned the open-source weight model Qwen3.5-9B using proprietary internal data, achieving better performance in the specialized task of anti-fraud compared to cutting-edge models, while being more cost-effective and faster. According to Coinbase's tech blog, the model's F1 score is 9.60% higher than that of cutting-edge models, and end-to-end latency has been reduced by 55.00%. Armstrong noted that cutting-edge models are excellent generalists, but specialized models will excel in more segmented tasks. Coinbase is enabling employees to easily fine-tune models internally, and some backend services may eventually be fully replaced by this new architecture.