deepseek-v4.1IMAGE

Arena: V4.1-Flash-20261017 first among open models on WebDev, 1670 overall

Citing Arena's official leaderboard, V4.1-Flash-20261017 takes first among open models on Code Arena WebDev ahead of Kimi K3 (Max) and ranks 12th overall at 1670 points, ahead of Claude Opus 5 (High) and 10 points behind GPT-6 Sol (Max), at $0.3/$1.2 per million tokens, about 10x cheaper.

AI-okww

@wenxiaohan99Archived 12 days agoOriginally posted 2026.10.19

Arena: V4.1-Flash-20261017 first among open models on WebDev, 1670 overall (image 1 of 2)

Arena: V4.1-Flash-20261017 first among open models on WebDev, 1670 overall (image 2 of 2)

AI-okww @wenxiaohan99Originally posted 2026.10.19

Open original on X

Arena 官方凌晨宣布：DeepSeek-V4.1-Flash-20261017 拿下 Code Arena: WebDev 开源榜第一，超过 Kimi K3 (Max)；总榜第 12，1670 分，排在 Claude Opus 5 (High) 前面，离 GPT-6 Sol (Max) 只差 10 分。重点是价格：输入 $0.3 / 输出 $1.2 每百万 token，比 Kimi K3 便宜约 10 倍。看第二张 Pareto 前沿图，它直接在图上打出一个新的拐点——同样的钱，没有比它更能打的了。上一场我刚推了它发布，这一场它就登榜，这迭代节奏是真不把人当对手。说真的，国产开源现在玩的已经不是"追赶"剧本了，是"性价比降维打击"——赌一个，Qwen 和 Kimi 这周必有回应。两张都是 Arena 官方原图，榜单那张建议存下来，下次有人问"开源模型到底行不行"直接甩图。

x.com/wenxiaohan99/status/2111991408311869828Archived 2026.10.19 12:02