Rohan Paul· @rohanpaul_ai · X·· 2026-08-13AI 评分43
AI 导读
Kimi K3 (Max) 在 Code Arena: WebDev 中以 1674 分领先 DeepSeek-V4-Pro (Max) 的 1607 分，但后者输入/输出价格分别便宜约 6.9 倍和 17.2 倍。
正文 · 原文
Kimi K3 (Max) leads DeepSeek-V4-Pro (Max) by 67 Arena points, but DeepSeek is roughly 6.9x cheaper on input and 17.2x cheaper on output.
Kimi $3 / $15 per MTok vs DeepSeek's $0.435/$0.87
For a simple 1M-input + 1M-output workload, that is $18 on Kimi versus $1.305 on DeepSeek, about 13.8x cheaper.
The 67-point difference is also an early AutoEval gap, not a linear percentage difference in coding quality, so Kimi earns that enormous premium only if it produces materially fewer failed edits, retries, or human reviews in your own environment.
Big news: DeepSeek-V4-Pro (Max) by @deepseek_ai is coming in around ~#8 overall (#2 among open models) in the Code Arena: WebDev! At 1607 pts, this places it after GPT-5.6 Sol (xHigh)(1622 pts), and makes it the second best open model after Kimi K3 (Max) (1674 pts). Note: this is an early AutoEval score, in which a Reward Model trained on Arena's human preference data casts automatic votes in place of live votes. We’ll continue to see how scores converge as more live human votes come in. See thread for Text Arena scores and more on AutoEval's methodology. Congrats to the @deepseek_ai team on this release!在 X 查看被引用的帖子
来源：Rohan Paul · x.com