Runpod report: Qwen has overtaken Meta’s Llama as the most-deployed self-hosted LLM

Runpod's State of AI report, based on anonymized serverless deployment logs from over 500,000 developers, reveals that Alibaba's Qwen has surpassed Meta's Llama as the most-deployed self-hosted LLM. Notably, Llama 4 shows near-zero adoption despite its launch hype, as developers prioritize performance per dollar, latency, and fine-tuning compatibility. For image generation, ComfyUI has become the de facto standard, powering over two-thirds of image endpoints. In AI video, optimization and upscaling workloads outnumber raw generation roughly two to one, indicating teams prefer iterative refinement over single expensive renders. Nearly two-thirds of Runpod's customers come from industries outside pure AI, with HealthTech and FinTech leading enterprise adoption.