OpenAI · LIVEBENCH 2026-06-25

GPT-5.4 Thinking xHigh Effort

Published benchmark results for this specific model configuration.

gpt-5.4-xhigh

Overall score78.0 / 100

Cost / successful task$0.387USD · benchmark workload

Weight accessNot reported

Where this configuration performs

Category averages from the same release, on a 0–100 scale.

Reasoning88.1

Coding77.5

Agentic Coding53.8

Mathematics94.1

Data Analysis79.3

Language82.6

Instruction Following70.2

Task-level results

theory of mind88.5zebra puzzle98.0spatial98.0logic with navigation68.0code generation74.6code completion80.4javascript68.2typescript43.3python50.0AMPS Hard98.0integrals with game93.0math comp94.1olympiad91.5consecutive events86.2tablejoin51.8tablereformat100.0connections100.0plot unscrambling65.9typos82.0paraphrase63.4simplify70.0story generation79.0summarize68.4

Source files: Scores ↗ Categories ↗ Costs ↗ . Missing cost is unknown, never free. Compare configurations within the same release.