OpenAI's GPT-6 Astra Outperforms GPT-5.6 Sol in Contracting Tasks
- Published
- Oct 6, 2026 — 10:00 UTC
GPT-6 Astra achieved an average score of 55.0% across 11 research tasks, outperforming GPT-5.6 Sol by 32%. The new model also reduced the estimated time per attempt to 19.2 minutes, which is 48% lower than the 37.0 minutes required by its predecessor. In terms of task criteria met, Astra achieved 94% compliance in approximately 20 minutes, compared to 85% for GPT-5.6 Sol in about 32 minutes. Sunita Verma, Chief Technology Officer at Ironclad, emphasized that effective AI in contracting must understand the entire lifecycle and business workflows. This advancement positions GPT-6 Astra as a more efficient tool for AI-driven contracting solutions, potentially enhancing productivity for teams relying on such technologies.
By Turing Wire Newsdesk · Oct 6, 2026 · How we work →
Summarised from OpenAI Blog's original report by the Turing Wire Newsdesk. Read the original for the full story.
Source: OpenAI Blog
