GPT-5.4 mini and GPT-5.4 nano, which can describe 76,000 photos for $52
Summary
OpenAI has introduced GPT-5.4 mini and GPT-5.4 nano, expanding its GPT-5.4 model series. The new GPT-5.4 nano model surpasses the previous GPT-5 mini in reasoning capabilities, while the new GPT-5.4 mini is twice as fast. These models feature highly competitive pricing, with GPT-5.4 nano being notably cheaper per million tokens than Google's Gemini 3.1 Flash-Lite. A practical demonstration showed GPT-5.4 nano describing a photo for 0.069 cents, projecting a cost of approximately \$52.44 to describe 76,000 images. The article also highlights the use of OpenAI Codex for generating images across various models and reasoning efforts, with the "gpt-5.4 xhigh" output being particularly noted for its quality.
Key takeaway
OpenAI's new GPT-5.4 mini and nano models deliver significantly more cost-effective and faster multimodal AI capabilities. GPT-5.4 nano outperforms the previous mini in reasoning and is cheaper than Gemini 3.1 Flash-Lite, enabling 76,000 image descriptions for just \$52. These models make large-scale image analysis and multimodal generation practical for budget-constrained production, though complex image generation quality can vary.
Topics
- GPT-5.4
- Large Language Models
- Multimodal AI
- AI Pricing
- Image Generation
Best for: CTO, Director of AI/ML, Computer Vision Engineer, AI Engineer, Machine Learning Engineer, AI Product Manager
Related on AIssential
Editorial summary, takeaway, and curation by AIssential. Original article published by Simon Willison's Weblog.