{"id":"gpt-5-4","slug":"gpt-5-4","name":"GPT-5.4","orgId":"openai","family":"gpt-5-4","releaseDate":"2026-03-05","availability":"ga","openWeights":false,"license":"Proprietary","paramsTotal":null,"contextWindow":1050000,"knowledgeCutoff":null,"modelCardUrl":"https://huggingface.co/datasets/lmarena-ai/leaderboard-dataset","hfRepo":null,"modalities":["text","image","file"],"aliases":["gpt-5.4-high","GPT-5.4","gpt-5.4-high (codex-harness)","gpt-5.4-medium (codex-harness)","gpt-5.4-2026-03-05_medium","gpt-5.4-2026-03-05_none","gpt-5.4-2026-03-05_high","gpt-5.4-2026-03-05_low","gpt-5.4-2026-03-05_xhigh","gpt-5.4-2026-03-05_unknown","openai/gpt-5.4","OpenAI: GPT-5.4"],"aliasNote":null,"organization":{"id":"openai","name":"OpenAI","slug":"openai","country":null,"url":null,"color":"#10A37F","mark":"O"},"price":{"modelId":"gpt-5-4","provider":"openai","inputPerM":null,"outputPerM":null,"blendedPerM":5.625,"tokensPerS":null,"observedAt":"2026-10-14","sourceId":"openrouter"},"answer":{"word":"It used to be.","scope":"Was SOTA in General LLM.","lead":"As of Oct 14, 2026, GPT-5.4 from OpenAI. Was SOTA in General LLM. Was SOTA in Factuality from 2026-08-10 to 2026-08-21. Currently ranked #6.","confidence":"medium"},"verdicts":[{"id":"v_overall_general-llm_gpt-5-4_2026-04-10_open_WAS_SOTA","model":"gpt-5-4","model_name":"GPT-5.4","category":"general-llm","category_name":"General LLM","slice":"overall","state":"WAS_SOTA","rank":28,"present_evidence":[],"missing_evidence":[],"margin":18.620689655172413,"confidence":"medium","valid_from":"2026-04-10","valid_to":null,"configuration":"high","rationale":"Was SOTA in General LLM from 2026-04-10 to 2026-04-20. Currently ranked #28.","safeguard_flags":["dethroned"],"co_sota_with":[],"challengers":[],"methodology":"1.0","label":"Was SOTA · General LLM"},{"id":"v_pareto_general-llm_gpt-5-4_2026-10-14_open_RANKED","model":"gpt-5-4","model_name":"GPT-5.4","category":"general-llm","category_name":"General LLM","slice":"pareto","state":"RANKED","rank":30,"present_evidence":[],"missing_evidence":[],"margin":15.957446808510639,"confidence":"medium","valid_from":"2026-10-14","valid_to":null,"configuration":"high","rationale":"Not on the General LLM cost-quality frontier.","safeguard_flags":[],"co_sota_with":[],"challengers":[],"methodology":"1.0","label":"#30 in General LLM"},{"id":"v_overall_reasoning_gpt-5-4_2026-10-14_open_RANKED","model":"gpt-5-4","model_name":"GPT-5.4","category":"reasoning","category_name":"Reasoning","slice":"overall","state":"RANKED","rank":29,"present_evidence":[],"missing_evidence":[],"margin":20.138888888888886,"confidence":"medium","valid_from":"2026-10-14","valid_to":null,"configuration":"high","rationale":"Ranked #29 in Reasoning.","safeguard_flags":[],"co_sota_with":[],"challengers":[],"methodology":"1.0","label":"#29 in Reasoning"},{"id":"v_pareto_reasoning_gpt-5-4_2026-10-14_open_RANKED","model":"gpt-5-4","model_name":"GPT-5.4","category":"reasoning","category_name":"Reasoning","slice":"pareto","state":"RANKED","rank":32,"present_evidence":[],"missing_evidence":[],"margin":17.55319148936171,"confidence":"medium","valid_from":"2026-10-14","valid_to":null,"configuration":"high","rationale":"Not on the Reasoning cost-quality frontier.","safeguard_flags":[],"co_sota_with":[],"challengers":[],"methodology":"1.0","label":"#32 in Reasoning"},{"id":"v_overall_math_gpt-5-4_2026-10-14_open_RANKED","model":"gpt-5-4","model_name":"GPT-5.4","category":"math","category_name":"Math","slice":"overall","state":"RANKED","rank":9,"present_evidence":[],"missing_evidence":[],"margin":17.777777777777757,"confidence":"medium","valid_from":"2026-10-14","valid_to":null,"configuration":"xhigh","rationale":"Ranked #9 in Math.","safeguard_flags":[],"co_sota_with":[],"challengers":[],"methodology":"1.0","label":"#9 in Math"},{"id":"v_pareto_math_gpt-5-4_2026-10-14_open_RANKED","model":"gpt-5-4","model_name":"GPT-5.4","category":"math","category_name":"Math","slice":"pareto","state":"RANKED","rank":9,"present_evidence":[],"missing_evidence":[],"margin":17.777777777777757,"confidence":"medium","valid_from":"2026-10-14","valid_to":null,"configuration":"xhigh","rationale":"Not on the Math cost-quality frontier.","safeguard_flags":[],"co_sota_with":[],"challengers":[],"methodology":"1.0","label":"#9 in Math"},{"id":"v_overall_science_gpt-5-4_2026-10-14_open_RANKED","model":"gpt-5-4","model_name":"GPT-5.4","category":"science","category_name":"Science","slice":"overall","state":"RANKED","rank":16,"present_evidence":[],"missing_evidence":[],"margin":null,"confidence":"low","valid_from":"2026-10-14","valid_to":null,"configuration":"xhigh","rationale":"Last listed at #16 in Science on 2026-03-05. That board does not award a crown.","safeguard_flags":["off_board"],"co_sota_with":[],"challengers":[],"methodology":"1.0","label":"#16 in Science"},{"id":"v_overall_coding_gpt-5-4_2026-10-14_open_INSUFFICIENT_DATA","model":"gpt-5-4","model_name":"GPT-5.4","category":"coding","category_name":"Coding","slice":"overall","state":"INSUFFICIENT_DATA","rank":null,"present_evidence":[{"benchmark":"swe-bench-verified","name":"SWE-bench Verified","rank":8}],"missing_evidence":[{"benchmark":"terminal-bench-4","name":"Terminal-Bench 4.0"},{"benchmark":"swe-bench-pro-hard","name":"SWE-bench Pro V2 HARD"},{"benchmark":"lmarena-text-coding","name":"LMArena Text (style-controlled, coding)"}],"margin":null,"confidence":"low","valid_from":"2026-10-14","valid_to":null,"configuration":"high","rationale":"#8 on SWE-bench Verified; missing Terminal-Bench 4.0, SWE-bench Pro V2 HARD, and LMArena Text (style-controlled, coding)","safeguard_flags":["off_board"],"co_sota_with":[],"challengers":[],"methodology":"1.0","label":"Not enough data"},{"id":"v_overall_coding-web-dev_gpt-5-4_2026-10-14_open_RANKED","model":"gpt-5-4","model_name":"GPT-5.4","category":"coding-web-dev","category_name":"Coding › Web dev","slice":"overall","state":"RANKED","rank":53,"present_evidence":[],"missing_evidence":[],"margin":49.523809523809526,"confidence":"medium","valid_from":"2026-10-14","valid_to":null,"configuration":"high","rationale":"Ranked #53 in Coding › Web dev.","safeguard_flags":[],"co_sota_with":[],"challengers":[],"methodology":"1.0","label":"#53 in Coding › Web dev"},{"id":"v_pareto_coding-web-dev_gpt-5-4_2026-10-14_open_RANKED","model":"gpt-5-4","model_name":"GPT-5.4","category":"coding-web-dev","category_name":"Coding › Web dev","slice":"pareto","state":"RANKED","rank":60,"present_evidence":[],"missing_evidence":[],"margin":47.2,"confidence":"medium","valid_from":"2026-10-14","valid_to":null,"configuration":"high","rationale":"Not on the Coding › Web dev cost-quality frontier.","safeguard_flags":[],"co_sota_with":[],"challengers":[],"methodology":"1.0","label":"#60 in Coding › Web dev"},{"id":"v_overall_long-context_gpt-5-4_2026-10-14_open_RANKED","model":"gpt-5-4","model_name":"GPT-5.4","category":"long-context","category_name":"Long context","slice":"overall","state":"RANKED","rank":33,"present_evidence":[],"missing_evidence":[],"margin":22.758620689655174,"confidence":"medium","valid_from":"2026-10-14","valid_to":null,"configuration":"high","rationale":"Ranked #33 in Long context.","safeguard_flags":[],"co_sota_with":[],"challengers":[],"methodology":"1.0","label":"#33 in Long context"},{"id":"v_pareto_long-context_gpt-5-4_2026-10-14_open_RANKED","model":"gpt-5-4","model_name":"GPT-5.4","category":"long-context","category_name":"Long context","slice":"pareto","state":"RANKED","rank":40,"present_evidence":[],"missing_evidence":[],"margin":21.276595744680847,"confidence":"medium","valid_from":"2026-10-14","valid_to":null,"configuration":"high","rationale":"Not on the Long context cost-quality frontier.","safeguard_flags":[],"co_sota_with":[],"challengers":[],"methodology":"1.0","label":"#40 in Long context"},{"id":"v_overall_vision_gpt-5-4_2026-10-14_open_INSUFFICIENT_DATA","model":"gpt-5-4","model_name":"GPT-5.4","category":"vision","category_name":"Vision","slice":"overall","state":"INSUFFICIENT_DATA","rank":null,"present_evidence":[{"benchmark":"lmarena-vision","name":"LMArena Vision (style-controlled)","rank":19}],"missing_evidence":[{"benchmark":"gdp-pdf","name":"GDP.pdf"}],"margin":null,"confidence":"low","valid_from":"2026-10-14","valid_to":null,"configuration":"high","rationale":"#19 on LMArena Vision (style-controlled); missing GDP.pdf","safeguard_flags":[],"co_sota_with":[],"challengers":[],"methodology":"1.0","label":"Not enough data"},{"id":"v_overall_factuality_gpt-5-4_2026-08-10_open_WAS_SOTA","model":"gpt-5-4","model_name":"GPT-5.4","category":"factuality","category_name":"Factuality","slice":"overall","state":"WAS_SOTA","rank":6,"present_evidence":[],"missing_evidence":[],"margin":4.545454545454561,"confidence":"medium","valid_from":"2026-08-10","valid_to":null,"configuration":"high","rationale":"Was SOTA in Factuality from 2026-08-10 to 2026-08-21. Currently ranked #6.","safeguard_flags":["dethroned"],"co_sota_with":[],"challengers":[],"methodology":"1.0","label":"Was SOTA · Factuality"},{"id":"v_pareto_factuality_gpt-5-4_2026-10-14_open_RANKED","model":"gpt-5-4","model_name":"GPT-5.4","category":"factuality","category_name":"Factuality","slice":"pareto","state":"RANKED","rank":6,"present_evidence":[],"missing_evidence":[],"margin":4.0697674418604635,"confidence":"medium","valid_from":"2026-10-14","valid_to":null,"configuration":"high","rationale":"Not on the Factuality cost-quality frontier.","safeguard_flags":[],"co_sota_with":[],"challengers":[],"methodology":"1.0","label":"#6 in Factuality"},{"id":"v_overall_arc_gpt-5-4_2026-10-14_open_RANKED","model":"gpt-5-4","model_name":"GPT-5.4","category":"arc","category_name":"Abstraction (ARC)","slice":"overall","state":"RANKED","rank":10,"present_evidence":[],"missing_evidence":[],"margin":null,"confidence":"low","valid_from":"2026-10-14","valid_to":null,"configuration":"xhigh","rationale":"Last listed at #10 in Abstraction (ARC) on 2026-03-05. That board does not award a crown.","safeguard_flags":["off_board"],"co_sota_with":[],"challengers":[],"methodology":"1.0","label":"#10 in Abstraction (ARC)"}]}