OpenAI推出GPT-6.1 Sol,在代码编写、文档理解等任务上显著优于前代,同时宣称成本更低。
TechCrunch reports GPT-6.1 Sol delivers measurable gains over GPT-6 Sol across professional workloads—including code generation, debugging, and document comprehension—while approaching the capability of the unreleased GPT-6 Astra. Crucially, OpenAI explicitly states it ‘costs less’, aligning with MIT Technology Review’s broader critique that many teams over-provision models unnecessarily. For early-stage AI founders, this validates a pragmatic path: prioritize validated, production-ready models with documented cost-performance tradeoffs over chasing bleeding-edge benchmarks. GPT-6.1 Sol isn’t just another version—it’s a signal that infrastructure maturity (e.g., optimized inference, caching, agent sandboxing in /) now matters more than raw parameter count for shipping real applications.