LLM Reference

GPT-5.4 nano

ProprietaryMultimodal

About

Ultra-efficient nano variant of GPT-5.4 for high-throughput, speed-optimized inference. Most cost-efficient GPT-5.4 model at $0.20 input / $1.25 output per 1M tokens. Released March 18, 2026.

Capabilities

MultimodalFunction CallingTool UseJSON Mode

Specifications

FamilyGPT-5
Released2026-03-18
Context400K
ArchitectureDecoder Only
Knowledge cutoff2025-08
Specializationgeneral
LicenseProprietary