OpenAIs GPT-5.4 mini and nano launch - with near flagship performance at much lower cost
Summary
OpenAI released GPT-5.4 mini and GPT-5.4 nano, smaller high-throughput models that deliver near-flagship GPT-5.4 performance while running much faster. Benchmarks show GPT-5.4 mini approaches full GPT-5.4 on multiple tests (e.g., SWE-bench Pro, Terminal-Bench 2.0, GPQA) and significantly outperforms the earlier GPT-5 mini; nano targets classification, extraction, ranking, and simpler coding tasks. The models target agentic, coding, and multimodal workflows and let developers mix large planning models with cheaper subagents to lower latency and cost. OpenAI made mini available via API, Codex, and ChatGPT with substantially lower pricing than GPT-5.4, while nano is API-only and priced even lower, and early customers report strong cost-performance benefits in real-world apps.
