XNews

Shopify CEO: Fine-Tuned Small Models Beat GPT 5.6 on Specialized Tasks

Shopify CEO Tobi stated that small models trained for specific purposes perform exceptionally well when paired with a self-improving recursive flywheel. The Shopify ML team fine-tuned Qwen3.5-0.8B and it outperformed GPT 5.6-sol on a very specialized task.

Source posts