三模型跑火柴人测试:DeepSeek V4.1 Flash 在 Opencode 上偏慢但便宜
作者用火柴人测试对比 Elorian、DeepSeek V4.1(1071 checkpoint,Opencode)和 Argon 4,认为 DeepSeek 速度不如预期但价格很低。
Lets compare this weeks new models! Above we have the new @ElorianAI which despite being made to analyze video and images, can also complete the Stickman test. In the middle we have Deepseek V4.1 with its new 1071 checkpoint on Opencode, which I though would be faster to be honest. But at least it is very cheap. And below Argon 4 for comparison, which @antigravity needs desperately to improve the usage on their plans. Elorian surrised me most here, considering it is a tiny fast model. Share your tests in the comments and let me know what you think of each of the 3 models so far.