GPT-5.4 nano is the smallest and most economical tier in OpenAI's GPT-5.4 lineup, announced on March 17, 2026 alongside GPT-5.4 mini as a fast, efficient alternative to the flagship model. Press coverage describes the model as inheriting many of the strengths of the larger GPT-5.4 while being tuned for coding assistance, subagent orchestration, and other high-volume workloads where reduced cost and quick turnaround matter more than peak capability. Its multimodal design accepts both text and image inputs while producing text-only outputs, giving it enough flexibility for grounded chat, document analysis, and tool-mediated pipelines without the overhead of a larger model.

For developers routing requests through the LLM Gateway, GPT-5.4 nano surfaces meaningfully different behavior depending on the upstream provider. OpenRouter's per-provider telemetry shows OpenAI direct delivering the fastest response, with a P50 latency of 0.64 seconds at roughly 59 tokens per second, while Azure and Azure US trade speed for higher availability at 99.75% and 99.95% uptime respectively, and an OpenAI Flex tier offers a separate throughput profile. That spread makes the model well suited to production scenarios where teams can choose between raw speed and steady reliability, particularly for embedded assistants, batch summarization, and lightweight agent loops where the the cataloged API limit context window comfortably accommodates long conversational histories or document collections.