Agent 評測
GAIA Agent 評測
OpenGPT 將15筆公開記錄保留在下方所示的 GAIA 評測版本與環境內。
- 工作
- 165
- 公開記錄
- 15
- 來源資料日期
- 身分核驗日期
- 歷史記錄
資料範圍
HAL verified snapshot · Top 15 published configurations · public validation set · 165 questions. public validation questions. General-assistant questions requiring reasoning, multimodality, browsing, and tools. Source-specific; no normalized step or token limit published.
OpenGPT 如何處理這些資料
OpenGPT 保留發布方的指標、工作範圍、環境、版本、設定欄位與資料日期,不產生跨評測 Agent 總分。
可比範圍
- 更高結果只適用於所展示的評測版本與條件。
- Agent 結果反映完整系統,而不是基礎模型單獨能力。
- 成本、穩定性、安全性或速度沒有公開時保持未報告,不進行推測。
- HAL reproduced these public-validation rows and publishes scaffold, model, score, cost, run count, and traces. The source has paused adding new models, and it does not publish a normalized tool or step budget.
HAL verified snapshot · Top 15 published configurations · public validation set · 165 questions
展示15筆公開記錄中的15筆;每項結果都保留來源評測與設定。
- HAL Generalist Agent · Claude Sonnet 4.5 (September 2025)Holistic Agent Leaderboard · Claude Sonnet 4.5 (September 2025)#1Accuracy: 74.55%HAL Generalist Agent
- HAL Generalist Agent · Claude Sonnet 4.5 High (September 2025)Holistic Agent Leaderboard · Claude Sonnet 4.5 High (September 2025)#2Accuracy: 70.91%HAL Generalist Agent
- HAL Generalist Agent · Claude Opus 4.1 High (August 2025)Holistic Agent Leaderboard · Claude Opus 4.1 High (August 2025)#3Accuracy: 68.48%HAL Generalist Agent
- HAL Generalist Agent · Claude Opus 4 High (May 2025)Holistic Agent Leaderboard · Claude Opus 4 High (May 2025)#4Accuracy: 64.85%HAL Generalist Agent
- HAL Generalist Agent · Claude-3.7 Sonnet High (February 2025)Holistic Agent Leaderboard · Claude-3.7 Sonnet High (February 2025)#5Accuracy: 64.24%HAL Generalist Agent
- HAL Generalist Agent · Claude Opus 4.1 (August 2025)Holistic Agent Leaderboard · Claude Opus 4.1 (August 2025)#6Accuracy: 64.24%HAL Generalist Agent
- HF Open Deep Research · GPT-5 Medium (August 2025)Hugging Face · GPT-5 Medium (August 2025)#7Accuracy: 62.80%HF Open Deep Research
- HAL Generalist Agent · GPT-5 Medium (August 2025)Holistic Agent Leaderboard · GPT-5 Medium (August 2025)#8Accuracy: 59.39%HAL Generalist Agent
- HAL Generalist Agent · o4-mini Low (April 2025)Holistic Agent Leaderboard · o4-mini Low (April 2025)#9Accuracy: 58.18%HAL Generalist Agent
- HF Open Deep Research · Claude Opus 4 (May 2025)Hugging Face · Claude Opus 4 (May 2025)#10Accuracy: 57.58%HF Open Deep Research
- HAL Generalist Agent · Claude-3.7 Sonnet (February 2025)Holistic Agent Leaderboard · Claude-3.7 Sonnet (February 2025)#11Accuracy: 56.36%HAL Generalist Agent
- HAL Generalist Agent · Claude Haiku 4.5 (October 2025)Holistic Agent Leaderboard · Claude Haiku 4.5 (October 2025)#12Accuracy: 56.36%HAL Generalist Agent
- HF Open Deep Research · o4-mini High (April 2025)Hugging Face · o4-mini High (April 2025)#13Accuracy: 55.76%HF Open Deep Research
- HAL Generalist Agent · o4-mini High (April 2025)Holistic Agent Leaderboard · o4-mini High (April 2025)#14Accuracy: 54.55%HAL Generalist Agent
- HF Open Deep Research · GPT-4.1 (April 2025)Hugging Face · GPT-4.1 (April 2025)#15Accuracy: 50.30%HF Open Deep Research