At its annual DevDay 2026 conference in San Francisco, OpenAI officially introduced its new GPT-6.1 Sol model. The launch represents a deliberate strategic move to make high-end model capabilities commercially viable for a wider range of developers. Priced well below the premier GPT-6 Astra model, Sol is designed to deliver comparable capabilities across core benchmarks. The release directly addresses intensifying competition across the enterprise software and interactive development sectors.
The pricing model for GPT-6.1 Sol is set at 2.00 US dollars per million input tokens and 10.00 US dollars per million output tokens. As a result, operating expenses represent only one-fifth of the cost required for GPT-6 Astra. Despite this substantial discount, the model achieves a score of 97.5 percent on established benchmarks such as DeepSWE v1.1 and KingBench 3. In both automated code generation and computer-use tasks, Sol approaches the capabilities of current flagship offerings.
Beyond benchmark statistics, real-world demonstrations for the creative and gaming industries served as the focal point of the presentation. Engineers demonstrated the rapid creation of playable 3D prototypes, SVG vector graphics, and interactive game mechanics directly generated by the model. The reduced latency and increased precision when translating natural language prompts into working software code mark a notable evolution. OpenAI is directly positioning this model for integration into digital media production pipelines.
Another major highlight of the conference was the unveiling of Dots. These are persistent, autonomous agents equipped with their own dedicated cloud environments. Dots are specifically designed to continue tasks and background routines autonomously, even when the human developer is offline or disconnected from the platform. This framework aims to bridge the gap between reactive conversational assistants and continuous background software agents.
These product launches arrive at a time of heightened rivalry among top-tier foundation model providers. Software developers immediately began comparing GPT-6.1 Sol against Claude Sonnet and Opus 5.5, evaluating their respective stability in agent workflows, 3D asset generation, and visual motion design. With competitors such as Anthropic and Meta accelerating their enterprise-focused initiatives, OpenAI is relying on cost-effective performance to safeguard its market footprint.