OpenAI: GPT-5.4

GPT-5.4 is a paid model from OpenAI that accepts text, images, and files as input. It supports tool use and reasoning, and offers a context window of 1,050,000 tokens with up to 128,000 tokens of output per response. Structured output support is unconfirmed from available data. The large context window suits long-document work, multi-turn agent loops, and tasks that require reasoning over substantial inputs. On benchmarks, GPT-5.4 scores 90.4 across four independent tests, with a perfect score on coding and a weaker 67.8 on agentic tasks, so its strengths are uneven. At $2.50 per million input tokens and $15.00 per million output tokens, the output cost is high relative to many alternatives. Teams doing heavy coding or reasoning work may find the performance justifies the price; those running high-volume pipelines or primarily agentic workflows should compare costs carefully before committing.