OpenAI Launches GPT-6.1 Sol with Astra-Level Intelligence at Scale

Ender Öztürk

September 29, 2026

OpenAI has officially unveiled its latest innovation, the GPT-6.1 Sol model, designed to redefine the equilibrium between high-end artificial intelligence performance and operational affordability. Announced today, this new model delivers intelligence comparable to the flagship GPT-6 Astra, yet it is available at one-fifth of the standard token price for both input and output operations. By lowering the barrier to entry for advanced agentic workflows and complex coding tasks, OpenAI aims to empower developers and enterprises to build highly capable autonomous systems without the prohibitive costs associated with previous top-tier models. This release marks a strategic shift toward making premium-grade AI accessible for large-scale professional applications.

The introduction of GPT-6.1 Sol addresses a critical need within the software development community for efficient, agentic coding tools. As companies increase their reliance on AI to handle multi-step workflows and complex debugging, the financial burden of large-scale token usage has become a primary concern. The Sol model mitigates this by providing a high-performance alternative that maintains accuracy in factual retrieval and reasoning tasks. By utilizing advanced context caching, the model allows for repeat requests at a 95 percent discount compared to standard input rates, making it an ideal candidate for long-running autonomous agents that require constant reference to large datasets.

Performance Benchmarks Prove High Capabilities

Technical evaluations indicate that GPT-6.1 Sol performs exceptionally well against its predecessors and competitors. In the rigorous DeepSWE v1.1 testing suite, the model achieved parity with GPT-6 Astra, despite operating at a significantly lower cost structure. Furthermore, the model demonstrated superior proficiency in analyzing intricate PDF documents, outperforming the Opus 5.5 benchmark. For tasks requiring computer interaction, the model surpassed its predecessor in OSWorld 2.0 testing, proving its utility in real-world professional environments.

Cost Efficiency Enhances Research and Development

Scientific research and data analysis workflows stand to benefit substantially from this release. By providing a 75 percent cost reduction for simulation and heavy data processing tasks, GPT-6.1 Sol offers a more sustainable path for academic and industrial research labs. The model also features improved transparency regarding its limitations and exhibits higher reliability in adhering to safety guidelines and user intent compared to earlier versions. This combination of increased accuracy and reduced overhead ensures that organizations can scale their AI-driven research without compromising on quality.

Availability Extends to All Professional Tiers

Starting today, the model is available for users subscribed to ChatGPT Plus, Pro, Business, Enterprise, and Edu tiers. For developers integrating the technology into their own infrastructure, the model is accessible through the OpenAI API under the identifier ‘gpt-6.1-sol’. Standard API pricing is set at 2 dollars per million input tokens and 10 dollars per million output tokens. Additionally, the launch includes ‘Ultrafast’ variants for both the Sol and Astra models, catering to high-frequency applications that prioritize low latency. A new Pro subscription tier, priced at 500 dollars per month, has been introduced to support users with the highest usage requirements.

We are eager to hear how the release of GPT-6.1 Sol might change the way you deploy AI agents in your own projects. Do you believe this aggressive pricing strategy will accelerate the adoption of autonomous coding tools in your industry? Share your thoughts and predictions in the comments section below.