gpt-6VIDEO
GPT-6.1-Sol tested against Astra at high reasoning effort
A user tests GPT-6.1-Sol with their bear-on-bicycle prompt and finds it clearly better than GPT-6-Astra at low reasoning effort when set to High, at one-fifth the price, and plans to use it for Computer Use scheduled tasks to save tokens.
李不凯正在研究
@libukaiArchived 28 days agoOriginally posted 2026.09.30
李不凯正在研究 @libukaiOriginally posted 2026.09.30
帮你们用我最爱的小熊骑车提示词试了一下，GPT-6.1-Sol 还真有点东西啊。 在思考强度调整到 High 时，要比 Low 思考的 GPT-6-Astra 好的多，就更不用说已经完全放飞自我的前代 GPT-6-Sol 了，😂 考虑到 5x 的价格差异，GPT-6.1-Sol 的确是性价比超高。 例如我那一大堆 Computer Use 定时任务，改用 GPT-6.1-Sol 应该能省下不少 Token 来。
This post quotes another post: x.com/libukai/status/2104985659488498013
x.com/libukai/status/2105002792507310460Archived 2026.09.30 10:20