GPT 6.1 Sol (openai.com)

OpenAI has unveiled GPT-6.1 Sol, a cost-effective upgrade to its GPT-6 Sol model that delivers performance comparable to the more advanced GPT-6 Astra for agentic coding and professional tasks at a fraction of the cost. With cached input priced at just $0.10 per million tokens—95% less than standard pricing and 50% below GPT-6 Sol—GPT-6.1 Sol allows developers to create sophisticated AI agents that can effectively reuse context across multiple requests. This new model enhances capabilities across a variety of complex professional tasks, such as writing code, interpreting documents, and managing multi-step workflows, achieving higher accuracy and efficiency than its predecessor. Significantly, GPT-6.1 Sol achieves near-Astra performance on benchmarks like DeepSWE v1.1, GDP.pdf, and AutomationBench, all while reducing costs dramatically. For example, it achieves scores that exceed GPT-6 Sol's best by 6.4 percentage points and beats Opus 5.5 at lower reasoning efforts with less than half the cost per task. Furthermore, it shows notable improvements in factual accuracy and user alignment, ensuring better transparency and reliability than GPT-6 Sol. This model is tailored for developers and businesses looking to leverage powerful AI tools without excessive costs, making advanced AI capabilities more accessible than ever.