OpenAI Releases GPT-6.1 Sol: Near-Astra Coding and Computer Use at One-Fifth of Astra’s Token Price
What happened
OpenAI launched GPT-6.1 Sol on September 29, 2026, as an upgrade over its previous GPT-6 Sol model. It delivers close to Astra-level performance for agentic coding, computer interaction, and professional tasks but charges at only about one-fifth of Astra’s token prices. Input processing costs $2 per million tokens, output costs $10 per million tokens, and cached inputs drop dramatically to $0.10 per million tokens. The model is accessible via the OpenAI API, ChatGPT Work, and Codex platforms.
Why it matters
GPT-6.1 Sol sharpens OpenAI’s pricing edge while closing most of the performance gap with Astra’s high-end coding and computer-use AI. For developers and businesses that rely on agentic AI for automation or complex workflows, this changes the economics of deploying powerful AI agents at scale. The cheaper cached input pricing encourages repeated or iterative use without proportionally higher costs. This lets companies afford richer, more persistent state handling for AI agents that operate with greater context over time.
The availability on multiple OpenAI platforms means builders can integrate or transition smoothly. It pressures competitors relying on more expensive token schemes or slower improvements in coding proficiency. Since professional-grade performance now comes with drastically reduced input expense, smaller teams or startups with tighter budgets gain better access to advanced AI coding and automation capabilities.
What to watch next
The key next focus is how Astra responds on pricing or performance—whether it maintains superiority or reacts to OpenAI’s cost leapfrog. Watch for adoption signals among enterprise customers and developers choosing GPT-6.1 Sol over Astra models based on total cost of ownership, not just raw capability benchmarks.
Also, monitor OpenAI’s incremental API rollout and platform integration speed. The adoption curve will reveal how much demand exists for cheaper AI agents in professional environments. Finally, keep an eye on any emerging use cases unlocked by the much cheaper cached input token costs, which could shift how AI agents handle multi-turn or long-term interactions.
AI Quick Briefs Editorial Desk