GPT Astra · 发布历史
OpenAI GPT Astra 系列已收录 2 次发布和 49 个测试版本。
GPT Astra 命名系列,直接前代关系依据官方材料保留。
全部版本
- GPT-6.2 Astra18 项测试
- GPT-6 Astra34 项测试
已收录测试 · 49
- Agents' Last Exam · v1.1; current chart task snapshot and scoring details unspecified; not assumed to be the 2026-10-17 snapshotGPT-6.2 Astra
- Terminal-Bench-Science · 0.1GPT-6.2 Astra、GPT-6 Astra
- Artificial Analysis Intelligence Index · v5.0.1; nine evaluationsGPT-6.2 Astra
- FrontierCode · 2.0GPT-6.2 Astra
- LifeVal · Version and exact task set unspecified in release articleGPT-6.2 Astra
- Harvey LAB-AA · Artificial Analysis v1.1; 120-task held-out set; hallucination-gated and ungated all-pass ratesGPT-6.2 Astra
- GDPval · v2 internal; scoring rubric and exact task snapshot unspecifiedGPT-6.2 Astra
- Vals Science Index · v1GPT-6.2 Astra
- BenchCAD · 2.1 Core; tools and geometric metric not specified in this chartGPT-6.2 Astra
- AutomationBench · 1.0.6; 47 tools; chart-reported score, scoring rule unspecifiedGPT-6.2 Astra
- AA-3D · Version and leaderboard snapshot time unspecified in release articleGPT-6.2 Astra
- AA-AnalystAgent · Version unspecified; all five independent attempts correctGPT-6.2 Astra
- HealthBench Professional · 525 physician-authored conversationsGPT-6.2 Astra、GPT-6 Astra
- MentalHealthBench · Overall; 1,215 synthetic conversations; four independent responses per task; clinician-authored rubricsGPT-6.2 Astra
- GPQA Diamond · 198-question Diamond subsetGPT-6.2 Astra、GPT-6 Astra
- Epoch Capabilities Index · Pre-release ECI quoted in the 2026-11-17 GPT-6.2 system card; exact index snapshot/version unspecified; continually updatedGPT-6.2 Astra
- FrontierMath Erdős · 68 problems posed or studied by Erdős, open as of August 2026; exact evaluation revision unspecifiedGPT-6.2 Astra
- User-flagged factuality · Refreshed 1,000 de-identified conversations: 229 retained cases and 771 new cases; separate general factuality and reported-issue graders; not comparable with earlier system-card versionsGPT-6.2 Astra
- Agents’ Last Exam · 55 professional fields; exact task version unspecifiedGPT-6 Astra
- OSWorld · 2.0 v2026.08.08 offline subset; partial scoreGPT-6 Astra
- BenchCAD · Version unspecified; tool-assisted 3D reconstruction; geometric-overlap scoreGPT-6 Astra
- BrowseComp · 1,266 题;单代理与多代理按协议区分GPT-6 Astra
- OpenScore String Quartets · OMR-NED; version unspecifiedGPT-6 Astra
- Artificial Analysis Intelligence Index · v4.1.1; source table does not state unitGPT-6 Astra
- Terminal-Bench · 4.0GPT-6 Astra
- DeepSWE · v1.1GPT-6 Astra
- FrontierCode · v1.1 Extended setGPT-6 Astra
- FrontierCode · v1.1 Main setGPT-6 Astra
- Artificial Analysis Coding Agent Index · v1.4GPT-6 Astra
- FrontierMath · Tier 4 v2GPT-6 Astra
- Humanity's Last Exam · 2,500-question benchmarkGPT-6 Astra
- GeneBench Pro · v13; see official footnote in main articleGPT-6 Astra
- LifeSciBench · Gold v1GPT-6 Astra
- ARC-AGI-3 · Responses API harness with two settings changed to better match real-world performanceGPT-6 Astra
- ARC-AGI-2 · Version/subset unspecified in sourceGPT-6 Astra
- ARC-AGI-1 · Version/subset unspecified in sourceGPT-6 Astra
- Terminal-Bench Science · 0.1; chart-reported resolution rateGPT-6 Astra
- FrontierMath · Tier 4 v2; chart-reported accuracyGPT-6 Astra
- AutomationBench · Version unspecified; chart-reported accuracyGPT-6 Astra
- Agents’ Last Exam · 55 professional fields; chart-reported accuracy; exact task version unspecifiedGPT-6 Astra
- FrontierCode · 1.1 Extended; chart-reported accuracyGPT-6 Astra
- GeneBench Pro · v13; chart-reported accuracyGPT-6 Astra
- Mind2Web · Benchmark version unspecified; updated Codex harnessGPT-6 Astra
- HealthBench · 48,000+ rubric items across 5,000 multi-turn conversationsGPT-6 Astra
- Agents’ Last Exam · V1; 55 subfields; exact task snapshot unspecifiedGPT-6 Astra
- MMMU-Pro · Vision format explicitly labeled by the official comparison chart; task release/version unspecifiedGPT-6 Astra
- OSWorld · 2.1 offline set; task count and exact task release unspecified in the selected GPT-6.1 Terra article/chartGPT-6 Astra
- Factual error rate on difficult prompts · Internal error-inducing prompts in the GPT-6.1 Terra comparison article; exact dataset version, sample size and user-flagged provenance unspecified.GPT-6 Astra
- ScreenSpot-Pro · 1,581 expert-annotated tasks across 23 professional applications and 3 operating systemsGPT-6 Astra