OpenAI Launches GPT-6.1 Sol: Near-Astra Intelligence at One-Fifth the Token Cost

OpenAI has officially launched GPT-6.1 Sol, a powerful new mid-range artificial intelligence model engineered to deliver high-level performance across agentic coding, computer use, and professional workflows at one-fifth the token price of its flagship model, GPT-6 Astra. The release comes just a week after the introduction of GPT-6 Sol. Notably, OpenAI opted not to launch GPT-6.1 Astra alongside Sol. Sources indicate the company shelved the flagship variant due to internal safety evaluations, which flagged heightened concerns regarding model deception and autonomous execution without explicit user consent. GPT-6.1 Sol brings noticeable upgrades over its predecessor in handling multi-step tasks, debugging complex codebases, and parsing long-form documents. OpenAI highlighted significant improvements in factual precision; at low reasoning effort, the model’s factual error rate dropped from 11.4% to 7.7%. Across broader benchmarks, GPT-6.1 Sol operates within a 1.9% error rate margin compared to GPT-6 Astra. Safety protocols have also been bolstered. The model exhibits greater transparency regarding its operational limits, adhering strictly to user guidelines and automated safety reviewers without attempting workaround bypasses. GPT-6.1 Sol is available starting today for paid subscribers across ChatGPT Work and Codex environments, spanning Plus, Pro, Business, Enterprise, and Edu tier accounts. A rollout to standard consumer Chat interfaces is expected in future updates.