Arena 周报：Gemini 4 Carbon 登顶文本榜，GPT-6.1 Terra 进入 Agent 榜第八
This Week in the Arena: Gemini is back, and Terra got even more efficient.
Arena 公布本周榜单，Google DeepMind 的 Gemini 4 Carbon（High）在 Text Arena 排名第 1、Agent Arena 第 2、Code Arena: WebDev 第 4，并改变了三个竞技场的帕累托前沿。OpenAI 的 GPT-6.1 Terra（Max）以第 8 名进入 Agent Arena，净提升 +10.6%，每任务中位成本 $0.62。Microsoft AI 的 MAI-Image-2.7-Preview 在单图编辑竞技场首次亮相排第 3，比上一版本 2.6 的第 6 名高 45 分。Arena 同时为 Agent Mode 推出 Connectors，用户可在单个任务中组合 Vercel、Linear、GitHub 等工具完成更复杂的工作流。
推荐理由
AI 评分 70
Arena 一周榜单汇总了 Gemini 4 Carbon、GPT-6.1 Terra 等模型在多个竞技场的排名与成本变化，可对照各模型的位置与价格效率。
主题
正文
This Week in the Arena: Gemini is back, and Terra got even more efficient.
- Gemini 4 Carbon (High) by @GoogleDeepmind landed #1 in Text Arena, #2 in Agent Arena, and #4 in Code Arena: WebDev. It reshaped the Pareto frontier across all three arenas. See first impressions from @petergostev in thread below.
- GPT-6.1 Terra (Max) by @OpenAI entered Agent Arena at #8 with +10.6% net improvement at $0.62 median cost per task, reshaping the Pareto frontier and offering price efficiency.
- MAI-Image-2.7-Preview by @MicrosoftAI debuted in the Single Image Edit Arena at #3, +45 pts above its previous variant 2.6 at #6.
- We introduced Connectors to Agent Mode. Our global community of users can combine tools like Vercel, Linear, GitHub, and more in a single task to complete more complex workflows from start to finish. Details in the thread. https://t.co/obaBwSIpEC