GPT-5.4
OpenAI
Parameters
—
Architecture
—
Released
—
License
—
About
No detailed description available for this model.
Benchmark Scores
| Benchmark | Score | Date |
|---|---|---|
|
AgentWorldBench MCP
general_agent
|
100.00%
|
24.06.2026 |
|
AgentWorldBench Search
general_agent
|
96.47%
|
24.06.2026 |
|
AgentWorldBench Terminal
general_agent
|
68.74%
|
24.06.2026 |
|
AgentWorldBench SWE
general_agent
|
92.91%
|
24.06.2026 |
|
AgentWorldBench Android
general_agent
|
81.37%
|
24.06.2026 |
|
AgentWorldBench Web
general_agent
|
62.17%
|
24.06.2026 |
|
AgentWorldBench OS
general_agent
|
88.37%
|
24.06.2026 |
|
AgentWorldBench Overall
general_agent
|
96.35%
|
24.06.2026 |
|
HLE (with tools)
stem_reasoning
|
73.73%
|
23.08.2026 |
|
BrowseComp
general_agent
|
90.35%
|
23.08.2026 |
|
BrowseComp Agent Swarm
general_agent
|
78.40
|
23.08.2026 |
|
DeepSearch QA
general_agent
|
50.75%
|
23.08.2026 |
|
Toolathlon Verified
general_agent
|
53.49%
|
23.08.2026 |
|
MCPMark
general_agent
|
65.29%
|
23.08.2026 |
|
Claw-Eval Pass^3
coding_agent
|
77.75%
|
23.08.2026 |
|
Apex-Agents
general_agent
|
76.52%
|
23.08.2026 |
|
OSWorld-Verified
general_agent
|
80.08%
|
23.08.2026 |
|
Terminal-Bench 2.0
coding_agent
|
91.53%
|
23.08.2026 |
|
SWE-bench Pro
coding_agent
|
72.12%
|
23.08.2026 |
|
SciCode (subtask)
stem_reasoning
|
88.83%
|
23.08.2026 |
|
Humanity's Last Exam
stem_reasoning
|
71.40%
|
23.08.2026 |
|
AIME 26
stem_reasoning
|
100.00%
|
23.08.2026 |
|
HMMT Feb 26
stem_reasoning
|
100.00%
|
23.08.2026 |
|
IMOAnswerBench
stem_reasoning
|
97.73%
|
23.08.2026 |
|
GPQA Diamond
stem_reasoning
|
97.16%
|
23.08.2026 |
|
MMMU-Pro
vision_language
|
94.39%
|
23.08.2026 |
|
CharXiv (RQ)
document_understanding
|
66.40%
|
23.08.2026 |
|
MathVision
vision_language
|
88.80%
|
23.08.2026 |
|
BabyVision
vision_language
|
48.18%
|
23.08.2026 |
|
V-Star
vision_language
|
100.00%
|
23.08.2026 |
|
MMLU-Pro
knowledge
|
88.71%
|
26.06.2026 |
|
CodeForces
stem_reasoning
|
90.98%
|
26.06.2026 |
|
Apex-Shortlist (no tools)
stem_reasoning
|
80.26%
|
26.06.2026 |
|
GDPVal-AA v2
general_agent
|
91.43%
|
26.06.2026 |
|
MCP-Atlas
general_agent
|
74.69%
|
26.06.2026 |
|
Claw-Eval Avg
coding_agent
|
94.98%
|
23.08.2026 |