OpenAI’s newest GPT-6 model is not simply chasing the highest benchmark score. GPT-6.1 Sol is aimed at a more practical target: bringing near-frontier capability to coding, computer use and business workflows at a dramatically lower operating cost.
For businesses building AI agents and automations, the important story is the price-to-performance shift. More capable workflows may become economical to run continuously rather than only for high-value tasks.
What OpenAI announced
OpenAI introduced GPT-6.1 Sol on 29 September 2026 as an upgrade to GPT-6 Sol. The company says the model approaches the performance of its higher-end GPT-6 Astra system across agentic coding, computer use and professional knowledge work, while standard API input and output pricing is one-fifth of Astra’s.
OpenAI lists GPT-6.1 Sol at US$2 per million input tokens and US$10 per million output tokens, with cached input at US$0.10 per million tokens. For developers running agents that repeatedly reuse company context, procedures or documents, that cached-input price is particularly significant.
Stronger at multi-step work
The bigger change is not just raw chat quality. OpenAI is positioning GPT-6.1 Sol for tasks where an AI system must work through several steps, use tools and verify its progress.
- Coding: OpenAI reports that GPT-6.1 Sol nearly matches Astra on its DeepSWE software-engineering evaluation while costing substantially less to run.
- Business workflows: On AutomationBench, which tests end-to-end work across areas such as sales, marketing, support, finance and HR, OpenAI reports an improvement over GPT-6 Sol.
- Computer use: The model is also stronger at interacting with software interfaces and completing longer workflows.
- Factuality: OpenAI says the model reduced factual errors on a deliberately difficult internal evaluation compared with GPT-6 Sol.
As with all vendor benchmarks, those figures should be treated as one part of the picture rather than proof that one model will be best for every company. Real workflows still need testing with the documents, systems and edge cases the business actually uses.
What this could mean for SMEs
The most interesting impact may be on the economics of AI automation. A business does not necessarily need the most powerful model available for every step. A well-designed workflow can use a strong, cost-efficient model for routine reasoning and reserve more expensive models for exceptional cases.
That matters for practical systems such as AI receptionists, enquiry qualification, proposal preparation, document processing, internal support assistants and workflows that connect websites, CRMs, email and business systems. When the cost of capable reasoning falls, more of those workflows become realistic to run at scale.
The real opportunity is not “having GPT”. It is deciding where AI should sit inside an existing process, what data it can access, what it is allowed to do and when a human should take over. Model improvements make the engine better; workflow design determines whether the business gets value from it.
Availability
At launch, OpenAI said GPT-6.1 Sol was available to developers through the API and to eligible ChatGPT users in Work and Codex, with broader product availability depending on plan and product. Businesses considering deployment should check current availability and pricing before designing a production workflow.
