Event ·Multi-source

GPT-6 Astra 在 ARC-AGI-3 基准测试中登顶，同时降低 token 成本

GPT-6 Astra 在 ARC-AGI-3 基准测试中取得第一，与 Claude Fable 5.1 在人工智能指数上持平。它支持异步工具调用并行执行，并保留了推理记忆。这些特性对开发者的代理应用至关重要。

Current read

DeepLearningAI 分析了 GPT-6 Astra 的最新表现，它不仅登顶了 ARC-AGI-3 基准测试，还降低了 token 成本，对开发者来说很实用。

11 sources78° AI heatLast updated 2026/9/14 15:31:54

Evidence

11 sources

Primary sourceDeepLearning.AI

2026/9/14 15:31:54View source

Related source 1lmarena.ai

2026/9/12 16:41:56View source

Related source 2Decoder

2026/9/13 10:52:16View source

Related source 3OpenAI Blog

2026/9/14 00:00:00View source

Related source 4AI Will

2026/9/14 04:12:29View source

Related source 5elvis

2026/9/14 12:48:01View source

Related source 6Gary Marcus

2026/9/14 13:09:08View source

Related source 7a16z

2026/9/14 22:15:16View source

Related source 8e27

2026/9/15 02:30:13View source

Related source 9Simon Willison

2026/9/13 00:19:32View source

Related source 10歸藏(guizang.ai)

2026/9/14 02:17:13View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.

Open the story