deepseek-v4.1IMAGE
Kimi blog lists two different DeepSeek V4.1 Flash scores for the same test
Kimi K3.1 Pro's blog shows DeepSeek V4.1 Flash at 42.3 in a TB 4.0 table but 33.8 on the same page's cost chart, citing different sources.
eallon
@eallon81287Archived 4 days agoOriginally posted 2026.10.24
eallon @eallon81287Originally posted 2026.10.24
📊同一篇博客，一个模型两个分。 Kimi K3.1 Pro 博客里，TB 4.0 表格给 DeepSeek-V4.1-Flash-1017 记 42.3，注脚说取自它自家发布稿。 同一页那张分数对成本图上，它是 33.8，注脚写数据来自 Artificial Analysis。 你说该信哪个？
x.com/eallon81287/status/2113833361080799446Archived 2026.10.24 12:06