OpenAI launches GPT-6.1 Sol at one-fifth of GPT-6 Astra's token prices

GPT-6.1 Sol is positioned as a middle option between the existing GPT-6 Sol and the higher-end GPT-6 Astra, offering near-Astra capability at a fraction of the price. The release emphasizes cost efficiency alongside capability gains, particularly for agentic tasks that reuse context across requests. According to OpenAI, GPT-6.1 Sol outperforms Opus 5.5 with fallbacks on GDP.pdf at less than half the cost per task, and beats Opus 5.5 by 2.2 percentage points on AutomationBench at medium reasoning effort at roughly a third of the cost. It also makes progress on computer-use tasks, coming within 2.1 percentage points of Astra's score on OSWorld 2.0's offline set at roughly one-seventh the cost per task.

The update also addresses factual accuracy. On difficult prompts flagged by users, GPT-6.1 Sol reduces the share of responses containing a factual error from 11.4% to 7.7% at low reasoning effort compared with GPT-6 Sol. Across tested reasoning settings, its error rate remains within 1.9 percentage points of Astra's at less than one-fifth the cost per task. OpenAI notes these error-inducing conversations are not representative of typical usage.

The outcome hinges on whether developers adopt the new model for production agents, and how quickly OpenAI expands its availability. For now, the main trade-off is intelligence versus cost: Astra remains the top choice for the hardest scientific research tasks, while GPT-6.1 Sol aims to be the practical default for everyday agentic workflows. The forthcoming GPT-6.1 Sol Ultrafast, which promises up to 8x faster token generation in Codex, may further shift that calculus once it arrives.