DeepSeek-V4.1-Flash tops Code Arena WebDev open-source leaderboard
Arena said on October 19 that DeepSeek-V4.1-Flash-1017 (Max), released October 17, scored 1670 on the Code Arena: WebDev leaderboard to become the highest-ranked open-source model, ahead of Kimi K3 (Max) and other open models and also ahead of closed models including Muse Spark 1.3 (Max), Claude Opus 5 (High) and Grok 4.7 (xHigh), trailing GPT-6 Sol (Max) by 12 points. The report says the model scored above GPT-6 Astra, GPT-6.1 Sol and Claude Opus 5 on the DeepSWE v1.1 software engineering test, with average cost per task down about 60% from the previous version; peak-hour pricing is 2 yuan per million input tokens on a cache miss and 8 yuan for output, falling to 0.04 yuan for cached input. DeepSeek says the model supports "wish-style programming" that generates games and 3D modeling in one click, and one netizen's test reported faster generation than GPT-6.1's ultra-fast mode.
According to Arena's comparison chart, the model enters Code Arena's "Pareto frontier," which the report says means no other model is currently both cheaper and higher-scoring.
From the digest “DeepSeek 新模型登顶开源模型榜首；国内 55% 算力用于视频和短剧？豆包发言人辟谣；微…”
DeepSeek 新模型登顶 Code Arena WebDev 开源模型榜首
10 月 19 日，AI 模型评测平台 Arena 宣布，DeepSeek-V4.1-Flash-1017（Max）超过 Kimi K3（Max），成为 Code Arena: WebDev 榜单排名最高的开源模型。
截至目前，DeepSeek-V4.1-Flash-1017 在该榜单上得分为 1670 分，超过了Kimi K3（Max）、Mimo-V2.6 Flash等开源模型，也超过了Muse Spark 1.3（Max）、Claude Opus 5（High）、Grok 4.7（xHigh）等闭源模型，仅落后 GPT-6 Sol（Max）12 分。
该模型于 10 月 17 日推出，在软件工程能力 DeepSWE v1.1 测试中，其得分超GPT-6 Astra、GPT-6.1 Sol、Claude Opus 5等模型，单项任务平均费用较上一版本下降约六成。DeepSeek 称其模型支持「许愿式编程」，可一键生成游戏、3D 建模等。网友实测该模型生成速度比 GPT-6.1 超快模式还要快。
价格方面，DeepSeek-V4.1-Flash-1017 在高峰时段的每百万 Token 输入（缓存未命中）价格为 2 元，输出价格为 8 元；空闲时段的输入（缓存未命中）价格为 1 元，输出价格为 4 元。若输入缓存命中，每百万 Token 输入价格在高峰时段为 0.04 元，空闲时段为 0.02 元，综合成本约为 Kimi K3 的 1/10。
Arena 发布的对比图显示，DeepSeek-V4.1-Flash-1017 进入了 Code Arena 的「帕累托前沿」，即图中的绿色折线，这说明目前没有其他模型能同时做到比它更便宜、且得分更高。 （来源：新智元）
#DeepSeek#Arena#Kimi K3#Claude Opus 5#GPT-6 Sol#Muse Spark 1.3#Grok 4.7#Mimo-V2.6 Flash