Google’s rollout of Gemini 4 Carbon on November 10 has driven its 80% implied probability, as the model quickly posted frontier results including a leading 59 on the Artificial Analysis Intelligence Index v5.0—edging Claude Opus 5.5—along with top marks in long-horizon agentic coding, cybersecurity benchmarks such as SEC-bench Pro, and practical tasks like CursorBench. Carbon builds directly on the earlier Gemini 4 Argon release with deeper reasoning and multi-agent capabilities, while remaining competitively priced during its introductory period. Traders view these confirmed capabilities and rapid availability as decisive for the November 23 resolution, leaving limited room for other frontier models to close the gap in the final days.
本市场将根据 arena.ai 的 Text Arena (Overall),在指定日期 12:00 ET 检查“Leaderboard”标签页下的表格时,排名最高的模型进行结算。 将使用 https://arena.ai/leaderboard/text/overall-no-style-control 处“Leaderboard”标签页“Text Arena | Overall”下“Rank”列的结果来结算本市场;样式控制已关闭(Adjustments: None),且筛选条件为“Models”。 注意:在相应检查时间标记为“AutoEval”的模型将不予考虑,无论其是否显示排名或分数。 市场创建后不会再添加新模型。本市场中未明确列出的任何模型都将归入“其他”选项。 模型主要按照市场检查时的排行榜排名排序。如果两个或更多模型排名并列,则按照其 Arena 分数排序,包括排行榜下方数据中反映的任何底层未四舍五入的精细数值。如果仍然平局,将以本市场组中列出的模型名称的字母顺序作为最终决胜依据(完整字符串,包括“-high”和“-max”等后缀);例如,如果两个模型仍然平局,“gemini-4-argon-high”的排名将高于“gemini-4-carbon-high”。本市场将根据此顺序由最先出现的模型结算。 本市场的结算来源是 arena.ai Text Arena (Overall)。如果检查时该结算来源不可用,本市场将保持开放,直到排行榜恢复在线,并根据其恢复可用后的首次检查结果结算。如果该来源永久不可用,本市场将结算为“其他”。
机器翻译。如有争议,以原文为准。