注册并分享邀请链接,可获得视频播放与邀请奖励。

Luke The Dev
@iamlukethedev
I'm a Hermes Builder - Hermes Bots Mobile App - Hermes Agora - Hermes Bible - Hermes 3D Christian • Dad • Life Lover -
加入 March 2009
274 正在关注    16.6K 粉丝
The model didn't change. The interfaces did. - same open 27B weights - 3.8x the success rate on OSWorld V2 - $1.10 a task vs $25.50 for GPT-5.5 and $72.40 for Opus 4.8 This is why I run Hermes with whatever model fits the job. The harness is half the game
显示更多
Our agent scientist, Agent S4, just got an open 27B model to beat GPT-5.5 and Claude Opus 4.8 on OSWorld V2, with the model's weights left exactly as they were. With the interfaces S4 evolved, Qwen3.8-27B completes 21.3% of the 108 tasks, the highest binary success of any open-source computer-use agent and 3.8x the same model with fixed interfaces. It costs $1.10 per task, against $25.50 for GPT-5.5 and $72.40 for Opus 4.8 💸 Full write-up:
显示更多