GPT 无后缀型号 · 发布历史
OpenAI GPT 无后缀型号 系列已收录 5 次发布和 75 个测试版本。
GPT 无后缀型号按版本分别记录;直接前代关系依据官方材料保留。
全部版本
已收录测试 · 75
- Terminal-Bench · 2.0;代理框架按协议区分GPT-5.5、GPT-5.4
- GDPval · 44 occupations; exact task/version unspecifiedGPT-5.5
- OSWorld-Verified · Verified benchmark; evaluation protocol noted per sourceGPT-5.5、GPT-5.4
- Toolathlon · Version, subset, and metric unspecified in the 2026-04-23 articleGPT-5.5
- BrowseComp · 1,266 题;单代理与多代理按协议区分GPT-5.5、GPT-5.4、GPT-5.2
- FrontierMath · Tier 1–3; version unspecifiedGPT-5.5
- FrontierMath · Tier 4; version unspecifiedGPT-5.5
- SWE-Bench Pro · PublicGPT-5.5、GPT-5.4、GPT-5.2
- FinanceAgent · v1.1GPT-5.5、GPT-5.4
- OfficeQA · Pro 133-question subsetGPT-5.5
- MMMU-Pro · Standard and Vision formatsGPT-5.5、GPT-5
- MMMU-Pro · Tool-enabled protocol; exact version unspecified (gpt55-mmmu-pro-with-tools)GPT-5.5
- MCP-Atlas · 所选材料未明确子集GPT-5.5、GPT-5.4
- τ²-bench · Telecom sectionGPT-5.5、GPT-5.4、GPT-5.2、GPT-5.1、GPT-5
- GeneBench · 2026-04-23 report; 103 evaluations across 10 domainsGPT-5.5
- BixBench · Version/subset unspecifiedGPT-5.5
- GPQA Diamond · 198-question Diamond subsetGPT-5.5、GPT-5.4、GPT-5.2、GPT-5.1、GPT-5
- Humanity's Last Exam · 2,500-question benchmarkGPT-5.5、GPT-5.4、GPT-5.2、GPT-5
- Humanity’s Last Exam · Tool-enabled protocol; exact subset/version unspecifiedGPT-5.5
- ARC-AGI-1 · Verified; exact subset not specifiedGPT-5.5
- ARC-AGI-2 · ARC Prize Verified · semi-private setGPT-5.5
- Artificial Analysis Intelligence Index · Exact edition unspecified in chartGPT-5.5
- AA-LCR · Component of the Artificial Analysis Intelligence Index; version unspecifiedGPT-5.5
- AA-Omniscience · Component of the Artificial Analysis Intelligence Index; version unspecifiedGPT-5.5
- CritPt · Component of the Artificial Analysis Intelligence Index; version unspecifiedGPT-5.5
- GDPval-AA · Artificial Analysis leaderboardGPT-5.5
- IFBench · Component of the Artificial Analysis Intelligence Index; version unspecifiedGPT-5.5
- SciCode · Component of the Artificial Analysis Intelligence Index; version unspecifiedGPT-5.5
- Artificial Analysis Coding Index · Exact edition unspecified in articleGPT-5.5
- HealthBench · 48,000+ rubric items across 5,000 multi-turn conversationsGPT-5.5、GPT-5.4、GPT-5.2、GPT-5
- HealthBench Professional · 525 physician-authored conversationsGPT-5.5
- GDPval · 44 occupations; exact task set/version unspecified in release articlesGPT-5.4、GPT-5.2
- Toolathlon · Release article does not state benchmark version or scoring protocol (2026-03)GPT-5.4
- ARC-AGI-1 · Verified set; release article does not specify task countGPT-5.4、GPT-5.2
- ARC-AGI-2 · Verified;题集组成未在发布表中说明GPT-5.4、GPT-5.2
- OfficeQA · 题集与精确计分协议未注明GPT-5.4
- MMMU-Pro · Standard/Vision 等格式未在所选材料中注明GPT-5.4
- FrontierScience · Research;版本与题集未注明GPT-5.4
- FrontierMath · Tier 1–3; version not stated in 2025–2026 release articleGPT-5.4、GPT-5.2
- FrontierMath · Tier 4; version not stated in 2026-03 release articleGPT-5.4
- OmniDocBench · 版本未注明;归一化编辑距离GPT-5.4
- GraphWalks · BFS;0–128K 上下文;题数未注明GPT-5.4、GPT-5
- GraphWalks · BFS;256K–1M 上下文;题数未注明GPT-5.4
- GraphWalks · parents;0–128K 上下文;题数未注明GPT-5.4、GPT-5
- GraphWalks · parents;256K–1M 上下文;题数未注明GPT-5.4
- HealthBench Hard · System-card evaluation; exact dataset version unspecifiedGPT-5.4、GPT-5
- HealthBench Consensus · System-card evaluation; exact dataset version unspecifiedGPT-5.4、GPT-5
- SWE-bench Verified · Verified subset (500 problems)GPT-5.2、GPT-5.1
- AIME · 2025 editionGPT-5.2、GPT-5.1、GPT-5
- MMLU Language (0-shot) · OpenAI GPT-5 system card, 2025-08-07; MMLU test set translated by professional human translators; chain-of-thought prompting. Section 3.10 says 13 languages, but Table 10 lists 14 language rows.GPT-5.2、GPT-5
- MMMU · Single-attemptGPT-5.1、GPT-5
- FrontierMath · Set/version unspecified in GPT-5 and GPT-5.1 release tables; Python tool enabled.GPT-5.1、GPT-5
- tau2-bench · Airline categoryGPT-5.1、GPT-5
- τ²-bench · Retail sectionGPT-5.1、GPT-5
- BrowseComp Long Context · 128k contextGPT-5.1、GPT-5
- SWE-Lancer · IC Diamond setGPT-5.1、GPT-5
- PaperBench · 10-paper subset; each paper requires <10GB external dataGPT-5.1
- MLE-Bench · 30 competition subset selected from tasks with <50GB data and <10h reference durationGPT-5.1
- SWE-bench Verified · GPT-5 发布评测固定子集;500 题中排除 23 题,n=477GPT-5
- Aider Polyglot · Diff or diff-fenced edit formatGPT-5
- HMMT 2025 · 2025 edition; no-tools protocol from GPT-5 developer releaseGPT-5
- CharXiv Reasoning · Version/split not stated in selected sourceGPT-5
- Video-MMMU · Test subsetGPT-5
- ERQA · Version unspecified in GPT-5 developer releaseGPT-5
- Scale MultiChallenge · o3-mini grader protocol from GPT-5 developer releaseGPT-5
- Internal API instruction-following evaluation · Hard subset; exact set/version unspecifiedGPT-5
- COLLIE · Version unspecified in GPT-5 developer releaseGPT-5
- OpenAI MRCR · 2 needles; 128k contextGPT-5
- OpenAI MRCR · 2 needles; 256k contextGPT-5
- BrowseComp Long Context · 256k contextGPT-5
- VideoMME · Long subset, with subtitle categoryGPT-5
- LongFact-Concepts · GPT-5 developer release; no toolsGPT-5
- LongFact-Objects · GPT-5 developer release; no toolsGPT-5
- FActScore · GPT-5 developer release; no tools; hallucination-rate metricGPT-5
- SimpleQA · No search-enabled; report F1GPT-5