WHAT NEW IN X10.29 THU

X 上的 AI 合集档案 · 22 个合集 · 58,925 条

gpt-6IMAGE

GPT-6.1 Terra 跑分:Agents' Last Exam 60.0%,单任务 $1.01 对比 Astra

Terra 在 Agents' Last Exam 得 60.0%,略高于 GPT-6 Astra 的 59.3%,单任务成本 $1.01 对 $7.24;但 Terminal-Bench 4.0 仅 50.3%,编码是短板。

aipulsedaily
@aipulseda1ly1 天前收录原推发布于 2026.10.27
GPT-6.1 Terra 跑分:Agents' Last Exam 60.0%,单任务 $1.01 对比 Astra
aipulsedaily @aipulseda1ly原推发布于 2026.10.27
在 X 上打开原推

OpenAI just dropped GPT-6.1 Terra at $1 in and $5 out. Half the price of GPT-6.1 Sol and Claude Sonnet 5.5. On Agents' Last Exam, a computer use test, it scores 60.0% at max effort. GPT-6 Astra scores 59.3%. Terra does it for $1.01 a task, Astra costs $7.24. Claude Opus 5.5 still scores highest at 63.2%, but it costs $8.82 a task. The catch is coding. Terra hits 50.3% on Terminal-Bench 4.0. Sonnet 5.5 hits 70.6%. Its live now in ChatGPT and Codex for Plus, Pro, Enterprise and Edu, and in the API.

x.com/aipulseda1ly/status/2114782151921062302采集于 2026.10.27 10:09