gpt-6IMAGE
Astra Max Effort uses about 25x the tokens per task of V4.1 Flash
LiveBench data shows DeepSeek V4.1 Flash has the lowest token cost per completed task, while GPT-6 Astra at Max Effort costs roughly 25 times as much.
雪瑜
@xueyu1125Archived 2026.09.14Originally posted 2026.09.14
雪瑜 @xueyu1125Originally posted 2026.09.14
目前DeepSeek V4.1 Flash 目前在LiveBench 完成任务的token开销是最低的 GPT-6 Astra Max Effort 单任务完成开销大概是V4.1 Flash的25倍 这边一看DeepSeek V4.1 Flash 一个任务需要多次调整也不是不能接受🤣
x.com/xueyu1125/status/2099309153709650124Archived 2026.09.14 16:19