OpenAI has unveiled its latest AI model, GPT-6.1 Sol, which promises capabilities similar to its flagship model, GPT-6 Astra, but at a fraction of the cost. This new model is particularly noteworthy for its performance in complex tasks such as programming, document comprehension, and managing multifaceted business processes, all while being five times more affordable.
The pricing structure for GPT-6.1 Sol is significantly lower than that of GPT-6 Astra, with standard rates for input and output tokens set at $2 and $10, respectively. Additionally, cached input tokens are available for just $0.10 per million tokens, marking a 95% reduction in costs. This pricing strategy makes GPT-6.1 Sol especially advantageous for tasks requiring repeated context usage across multiple queries.
In direct comparisons, GPT-6.1 Sol outperformed its predecessor, GPT-6 Sol, in complex professional tasks, demonstrating superior capabilities in areas such as code writing and debugging. For instance, in the DeepSWE v1.1 test, which evaluates software development problem-solving within real codebases, GPT-6.1 Sol exceeded the performance of GPT-6 Sol by 6.4 percentage points while requiring less cognitive effort.
The new model also excelled in the GDP.pdf test, which assesses accuracy on professional inquiries related to complex PDF documents across various sectors, including finance and healthcare. Here, GPT-6.1 Sol outperformed competing models like Opus 5.5 by more than double, while maintaining lower operational costs.
Further tests, such as AutomationBench, highlighted GPT-6.1 Sol's effectiveness in executing multi-step business processes, achieving a 2.2 percentage point advantage over Opus 5.5 with significantly reduced costs. In tasks requiring interaction with computer applications, it notably surpassed GPT-6 Sol by 7 percentage points while remaining within range of Astra.
Overall, GPT-6.1 Sol not only shows improvements in factual accuracy with a lower error rate compared to GPT-6 Sol, but also demonstrates enhanced compliance with human intentions and values. This model is currently accessible to Plus, Pro, Business, Enterprise, and Edu users within the ChatGPT Work and Codex frameworks, although it is yet to be made available in the standard ChatGPT interface.
The introduction of GPT-6.1 Sol could significantly disrupt the AI market, offering a compelling alternative to existing high-cost models and prompting competitors to re-evaluate their pricing and service strategies.
Informational material. 18+.
