WHAT NEW IN X11.01 SUN

AI collections from X · 22 collections · 63,201 posts

deepseek-v4.1IMAGE

Bug Hunt test: DeepSeek V4.1 Flash scores 42.7 on 105 planted bugs, beating Opus 5.5 and Kimi K3.1 Pro

On an independent Bug Hunt benchmark counting only targeted fixes across 105 planted bugs in two production repos, V4.1 Flash 1017 scored 42.7 at $1–2, versus Kimi K3.1 Pro's 36 at nearly $14.

xiax|AI & Agents
@xiax0603Archived 7 days agoOriginally posted 2026.10.24
Bug Hunt test: DeepSeek V4.1 Flash scores 42.7 on 105 planted bugs, beating Opus 5.5 and Kimi K3.1 Pro
xiax|AI & Agents @xiax0603Originally posted 2026.10.24
Open original on X

💥反差拉满!Bug Hunt硬核测试,DeepSeek V4.1 Flash 1017显著强于Kimi K3.1 Pro 大家都在热议Kimi K3.1 Pro的3D、长文本与金融Agent能力,这份独立Bug Hunt基准却给出反差结果。测试采用两份生产仓库、105个植入漏洞,仅统计定向修复成绩,剔除意外修复带来的虚高。 DeepSeek V4.1 Flash 1017【MAX】拿到42.7分,超过Opus 5.5和Kimi K3.1 Pro,成本仅1至2美元; Kimi K3.1 Pro【MAX】得分36,成本接近14美元,单修复成本差距悬殊。 没有全能的大模型。Kimi擅长3D创作、长链路Agent,但在存量代码深挖、隐蔽bug定位上存在短板。不同场景的最优模型本就不同。 你写代码,更看重从零搭建项目,还是排查存量bug?

This post quotes another post: x.com/xiax0603/status/2113831901110947985

x.com/xiax0603/status/2113837063330287741Archived 2026.10.24 12:05