elvis@omarsar0
36AI 编辑部评分,满分 100
2026-08-02 05:40· 45分钟前
跳到正文
AI 摘要

我一直说,这些模型的 token 效率被低估了。对这些模型要更大胆一些,尝试不同的 harness。 “智能便宜到无需计量”的时代已经到来! [引用 @cline]:虽然 DeepSeek V4-Flash 的每 token 价格显著更低,但如果因交互轮次更多导致单任务总成本更高,这一优势可能具有误导性。 不过,@ArtificialAnlys 报告称,DeepSeek 以比 Fable 低 105 倍的成本完成了相同的基准任务。

I keep saying that the token efficiency on these models are underestimated. Be more ambitious with these models. Try different harnesses.

Intelligence too cheap to meter is here!

ClineWhile DeepSeek V4-Flash is significantly cheaper on price per token, this can be misleading if the overall cost per task ends up being higher due to more turns ...
elvis · @omarsar0 · X·2026-08-02 05:40·45分钟前
在 X 看原推· x.com
AI 摘要

我一直说,这些模型的 token 效率被低估了。对这些模型要更大胆一些,尝试不同的 harness。 “智能便宜到无需计量”的时代已经到来! [引用 @cline]:虽然 DeepSeek V4-Flash 的每 token 价格显著更低,但如果因交互轮次更多导致单任务总成本更高,这一优势可能具有误导性。 不过,@ArtificialAnlys 报告称,DeepSeek 以比 Fable 低 105 倍的成本完成了相同的基准任务。

I keep saying that the token efficiency on these models are underestimated. Be more ambitious with these models. Try different harnesses.

Intelligence too cheap to meter is here!

ClineWhile DeepSeek V4-Flash is significantly cheaper on price per token, this can be misleading if the overall cost per task ends up being higher due to more turns ...
在 X 查看原推x.com