OpenAI announced the GPT-6.1 Sol model. Offering performance close to GPT-6 Astra, the model provides new opportunities for developers at one-fifth the cost.
OpenAI redefines the balance of cost and performance in the world of artificial intelligence by announcing the GPT-6.1 Sol model. The new model brings the high intelligence offered by GPT-6 Astra to users at one fifth of the standard input and output token prices.
This new version provides remarkable efficiency, especially in agent-based coding processes and professional workflows. Developers now have the opportunity to create more capable agents with more limited budgets.
High Performance and Low Cost with GPT-6.1 Sol
GPT-6.1 Sol is a strong performer for agent-based coding, computing, and professional work. The model, which comes close to GPT-6 Astra in challenging missions, provides developers with the flexibility to create more capable agents with the same budget.
The cost of cached entries is kept at just $0.10 per million tokens. This pricing means a 95 percent discount compared to the standard input price.
This creates a significant cost advantage for agent-based workloads that must reuse context across repeated requests. While the GPT-6 Astra remains the top-tier model in terms of overall capability, the Sol model strikes a new balance between cost and capability.
GPT-6.1 Sol offers an ideal option for important work that users want to do more frequently and for API customers who are developing large-scale applications. The model stands out as the most cost-effective option based on the level of performance it offers today.
Advanced Skills and Technical Assessments
GPT-6.1 Sol offers significant improvements over GPT-6 Sol in areas such as writing code, debugging, and executing multi-step workflows. In the DeepSWE v1.1 test, the model received the same score as the GPT-6 Astra, while its cost remained at approximately one-fifth.
It also scores higher than Opus 5.5 when it comes to analyzing complex PDF documents. In tasks requiring computer use, it manages to outperform GPT-6 Sol in the OSWorld 2.0 test.
It provides over 75 percent cost advantage over its competitors in scientific research processes, especially in areas such as data analysis and simulation. In terms of factual accuracy, GPT-6 provides more reliable results than Sol.
GPT-6.1 Sol offers a structure that is more transparent about its limitations. It also exhibits higher reliability than previous versions when it comes to complying with user intent and security constraints.
Access and New Subscription Options
The new model is available on ChatGPT Work and Codex platforms for Plus, Pro, Business, Enterprise and Edu users as of today. Developers can access the model via the OpenAI API under the name gpt-6.1-sol.
Standard API prices are set at $2 per million input tokens and $10 per million output tokens. In addition, GPT-6 Astra Ultrafast and GPT-6.1 Sol Ultrafast models, which stand out with their speed up to 8 times that of GPT-6 Astra, are also available via API.
The new Pro subscription level, offered for a monthly fee of $500, provides users with the highest usage limits. What do you think about the performance and cost advantage offered by these new models?
