openai.com
|
ksl
|
|
OpenAI released GPT-5.4 mini and nano, two smaller models built explicitly for high-volume, latency-sensitive workloads like subagents, coding assistants, and real-time image reasoning. Mini is over 2x faster than GPT-5 mini and approaches the flagship GPT-5.4 on benchmarks like SWE-Bench Pro, while nano drops to $0.20 per million input tokens – cheap enough to describe 76,000 photos for $52, as Simon Willison calculated. Mini goes to free ChatGPT users; nano is API-only. The framing around subagents and screenshot interpretation is deliberate – OpenAI is designing these models for the orchestration layer, not the chat window. Anthropic’s Haiku and Google’s Flash have been competing in this same cost-per-token bracket, and the margin for differentiation is getting thinner with every release.
