OpenAI and Ironclad partner on research into agents' computer use
What happened
On October 6 OpenAI disclosed a partnership with AI contract platform Ironclad, turning 11 legal and procurement tasks, including contract configuration and approval workflows, into training and evaluation tasks to push forward agents' computer use. The report says its first frontier model, GPT-6 Astra, scored 55.0% on internal evals, above GPT-5.6 Sol's 41.6%, and cut estimated time per attempt from 37.0 minutes to 19.2 minutes. What has been disclosed so far is the joint task setup and internal eval results.
Written by AI from the coverage · updated 37 minutes ago
Coverage
Follow the reports to see the story from different sides.
- OpenAI NewsAdvancing computer use with Ironclad
OpenAI 与 AI 合同平台 Ironclad 合作，将合同配置、审批流程等 11 项法律与采购任务转化为训练和评估任务，其首个前沿模型 GPT‑6 Astra 在内部评测中平均得分 55.0%，高于 GPT‑5.6 Sol 的 41.6%，单次尝试预估耗时从 37.0 分钟降至 19.2 分钟。
Heat over time
Not enough continuous observations to draw a trend yet.