OpenAI launches GPT‑5.4 Mini/Nano
OpenAI released GPT‑5.4 Mini and Nano — smaller, lower‑latency variants tuned for coding assistants, vision + tool use, and multi‑model/subagent workflows. They’re pitched for real‑time coding and high‑volume tasks with special pricing notes for European developers. (en.eloutput.com)
OpenAI published GPT‑5.4 mini and nano on March 17, 2026 and listed GPT‑5.4 mini as available in ChatGPT, Codex, and the public API. (openai.com) In OpenAI’s benchmark table, GPT‑5.4 scored 57.7% on SWE‑Bench Pro vs GPT‑5.4 mini at 54.4% and GPT‑5.4 nano at 52.4%, while OSWorld‑Verified shows 75.0% for GPT‑5.4, 72.1% for mini, and 39.0% for nano. (openai.com) The GPT‑5.4 mini API exposes a 400,000‑token context window, a 128,000‑token max output limit, and an Aug 31, 2025 knowledge cutoff while supporting text input and image input for text outputs. (developers.openai.com) OpenAI’s published unit prices for the new models list GPT‑5.4 mini at $0.75 per 1M input tokens and $4.50 per 1M output tokens, and GPT‑5.4 nano at $0.20 per 1M input tokens and $1.25 per 1M output tokens. (developers.openai.com) The platform pricing page and model docs state a 10% regional processing (data‑residency) uplift for GPT‑5.4, GPT‑5.4 mini, GPT‑5.4 nano and GPT‑5.4 pro when using hosted EU/regional endpoints. (developers.openai.com) OpenAI says GPT‑5.4 mini consumes roughly 30% of a GPT‑5.4 quota in Codex, and GitHub Copilot made GPT‑5.4 mini generally available in its model picker on March 17, 2026 with admin enablement and bring‑your‑own‑key options for organizations. (developers.openai.com) (github.blog) OpenAI included customer feedback noting that GPT‑5.4 mini “matched or exceeded competitive models on several output tasks” and reported higher end‑to‑end pass rates and stronger source attribution in early evaluations. (openai.com)
