Netlify partnered with Open Router to expand AI model choices through its AI Gateway and Agent Runners. Users can now access models like Kimi K3, GLM 5.2, and Deep Seek V4 for building projects. The platform's internal evaluation tool AXIS tests models using identical prompts across use cases: a coffee shop website, to-do list app, and recipe generator.

Testing revealed significant credit usage variations. Claude Opus averaged 551 credits but spiked to 1,055 credits in one run. GPT 5.6 Sol showed efficient performance at 141 credits average. Deep Seek V4 Flash proved most economical at just 1.3-2.5 credits per run.

For the coffee shop site test, models ranged from budget-friendly Deep Seek V4 Flash to premium Claude Opus. While Opus produced detailed designs with animated elements and dark mode support, the question remains whether its higher credit consumption justifies the visual output compared to more economical alternatives.