跳到正文
原文
The Decoder:AI News· Matthias Bastian·· 2 天前AI 评分53

Gemini 4 Carbon 登顶 MathArena 数学基准,成本约为 GPT-6.1 Sol 的 15 倍

Gemini 4 Carbon takes the MathArena crown but costs 15 times more than GPT-6.1 Sol

文章要点
  • 1Gemini 4 Carbon 登顶 MathArena 数学基准,成为近期首个占据榜首的非 OpenAI 模型。
  • 2测试所用推理设置下,Carbon 与 Claude Fable 5.2 的成本约为 GPT-6.1 Sol 的 15 倍和 40 倍。
  • 3Carbon 最常见的失败是超出时限,几乎从不放弃任务,曾运行一个超过 11 小时不查进度的脚本。
  • 4榜单前三预期得分为 85.4%、82.8%、80.6%,最佳开源模型 Kimi K3.1 Pro 为 54.3%。
决策要点

数学能力榜首与性价比已经脱钩,登顶的 Carbon 单次成本是第三名 GPT-6.1 Sol 的 15 倍,选型看榜单之外更要算账单。

  • 影响:数学密集型任务上榜首与第三名的精度差约 4.8 个百分点,成本差却达 15 倍,多数日常任务走低价档更划算;Carbon 几乎不放弃任务、曾跑 11 小时脚本,接入后超时和失控运行会直接放大推理支出。
  • 参考:可考虑把模型按任务难度分档,日常数学与推理任务先走低价档,高难题再小规模试 Carbon 并记录单题成本;接入时值得给长时运行加超时和进度检查上限,并盘点哪些场景真需要这个数学强度。
  • 依据:得分与成本为 MathArena 榜单的预期值口径,属第三方基准数据而非厂商宣传;失败模式与 11 小时脚本案例出自该基准测试的观察。
正文

Skip to content

Matthias Bastian

Nov 14, 2026

Google's Gemini 4 Carbon has become the first non-OpenAI model in a while to top the MathArena math benchmark. It knocked GPT-6.1 Sol from the top spot, while Claude Fable 5.2 came in second, making it potentially one of the most capable publicly available models for math tasks. OpenAI's models, which recently allegedly solved demanding equations like the Navier-Stokes equations, aren't publicly available yet.

Gemini 4 Carbon's most common failure mode is exceeding the time limit. The model also almost never gives up on a task, a trait that can be read as either positive or negative. In one case, it launched a script that ran for more than eleven hours without checking on its progress. That extra performance comes at a price, though. With the reasoning settings used in testing, Carbon and Fable 5.2 cost roughly 15 and 40 times as much as GPT-6.1 Sol, respectively.

Recommendation Model Provider Release date Expected performance Expected cost
#1 overall Gemini 4 Carbon (high) Google November 10, 2026 85.4% ±3.0% $15.70 ±$1.33
#2 overall Claude-Fable-5.2 (max) Anthropic November 9, 2026 82.8% ±5.3% $40.20 ±$2.59
#3 overall GPT-6.1 Sol (max) OpenAI September 29, 2026 80.6% ±2.7% $1.07 ±$0.018
Best open model Kimi K3.1 Pro (max) Moonshot AI October 23, 2026 54.3% ±5.8% $3.70 ±$0.27

AI News Without the Hype – Curated by Humans

Subscribe to THE DECODER for ad-free reading, a weekly AI newsletter, our exclusive "AI Radar" frontier report six times a year, full archive access, and access to our comment section.

来源:The Decoder:AI News · the-decoder.com