OpenAI has officially unveiled its new language model, GPT-6.1 Sol, which offers users and developers flagship-level programming and deep document processing capabilities while costing five times less than the previous top-tier Astra version. The creators position this algorithm as a breakthrough solution offering the market's best performance-to-cost ratio.
Cost-Efficiency and Performance
A key advantage of the new product is its pricing policy: cached input data for GPT-6.1 Sol costs a mere $0.10 per 1 million tokens. This is 95% cheaper than standard pricing, delivering massive savings when reusing context multiple times in complex agentic systems. Despite its democratic price tag, in DeepSWE v1.1 software development tests, the new model matched the flagship GPT-6 Astra, outperforming the previous GPT-6 Sol version by 6.4%.
Benchmark Results and Reliability
In the GDP.pdf benchmark for analyzing complex financial and legal documents, the model decisively outperformed its competitor Claude Opus 5.5. Furthermore, in the AutomationBench test for executing multi-step business processes, the system surpassed Opus 5.5 by 2.2% while consuming three times fewer resources. The developers also managed to reduce factual errors in complex prompts to 4.1%, coming within striking distance of the flagship Astra's 4% error rate.
Availability and Ultrafast Versions
The new neural network is already available to ChatGPT Plus, Pro, Business, Enterprise, and Edu subscribers in specialized Work and Codex workspaces, as well as via API under the identifier gpt-6.1-sol. Base pricing is set at $2 per 1 million input tokens and $10 per 1 million output tokens. Concurrently, OpenAI launched ultrafast modifications GPT-6 Astra Ultrafast and GPT-6.1 Sol Ultrafast, running up to 8 times faster than originals for premium Pro tier subscribers paying $500 per month.
