Shopify post-trains Nemotron 3.5 Lightning to slash inference costs by 90% and accelerate retail AI
Shopify post-trained NVIDIA Nemotron 3.5 Lightning for its GraphQL subagent, achieving 90% lower serving costs and 7x faster responses while surpassing leading frontier models in accuracy. NVIDIA CCluster vLLM optimization software further reduced inference costs up to 12x across five additional retail use cases. European retailers including Zooplus and Photoroom deployed NVIDIA models to improve search accuracy and product image processing, with listings selling at higher rates.
Shopify's 90% cost reduction and 7x speedup with Nemotron 3.5 Lightning demonstrates how post-training can make frontier models commercially viable, offering retailers a clear path to lower AI infrastructure costs and improved operational efficiency.
Source: NVIDIA Blog · blogs.nvidia.com