OpenAI launches GPT-5.4 mini and nano — faster small models built for coding, agents, and lightweight tasks

OpenAI released two new small models on March 17: GPT-5.4 mini and GPT-5.4 nano. GPT-5.4 mini is available immediately in ChatGPT, Codex, and the API — optimized for coding, computer use, multimodal understanding, and subagent orchestration, and clocked at more than 2x the speed of GPT-5 mini. GPT-5.4 nano is the smaller companion: OpenAI's smallest and cheapest GPT-5.4 variant, positioned for lightweight, high-volume tasks where cost and latency matter more than raw capability. Both models ship as the GPT-5.1 line reaches end-of-life — as of March 11, GPT-5.1 Instant, Thinking, and Pro have been retired and automatically migrated to GPT-5.3 and 5.4 equivalents.

The mini/nano naming is doing a lot of work here. OpenAI is clearly building a tiered model stack — full GPT-5 for general intelligence, GPT-5.4 Thinking for frontier reasoning, GPT-5.4 mini for agentic and coding workloads, GPT-5.4 nano for cost-optimized inference. That last tier is the telling one: nano signals OpenAI is competing directly for the embedded, high-throughput use cases where price-per-token is the buying criterion — territory currently dominated by Anthropic's Haiku and Google's Flash lines.