deepseek-v4.1POST
FrontierCode 合并率:Opus 5.5 56.0% vs V4.1 Flash 52.6%,单任务 $11.
在按「维护者是否会合并」评分的 FrontierCode 上,Claude Code 里的 Opus 5.5 得 56.0%,Codex 接 DeepSeek V4.1 Flash 得 52.6%,但单任务成本相差近 30 倍。
Farseen
@farseen5 天前收录原推发布于 2026.10.25
Farseen @farseen原推发布于 2026.10.25
on frontiercode , the part of the coding agent index graded on whether a maintainer would merge it , opus 5.5 in claude code scores 56.0% and codex with deepseek v4.1 flash scores 52.6% . per task that's $11.42 against $0.39
x.com/farseen/status/2114111932878008739采集于 2026.10.25 12:00