OpenAI’s GPT-6.1 Sol launch positions the model as an upgrade for coding, computer use and professional work, approaching GPT-6 Astra on selected evaluations at lower standard token prices. Its release page lists access through the API, ChatGPT Work and Codex, while distinguishing those products from Chat, where it says the model is not yet available.
The announcement makes cost-performance the central story. For a Canadian developer or business, the useful unit is a correctly completed task. Cheap input tokens do not help much if an agent needs repeated attempts, generates an unnecessarily long answer or leaves substantial repair work for a reviewer.
What the published comparisons establish
OpenAI reports improvements over GPT-6 Sol on software engineering, document-based professional questions and computer-use tasks. These are results from specified evaluation settings, not a guarantee that every application will improve by the same amount. The release also says benchmark environments can differ from production ChatGPT.
The company lists standard API prices of US$2 per million input tokens, US$0.10 for cached input and US$10 for output. Pricing should be verified before budgeting. Cached input is especially relevant to agents that reuse context, but the actual bill still depends on which requests qualify and how much output the workflow produces.
Related sources: OpenAI GPT-6.1 Sol release ↗
Build a comparison around work you can score
Choose representative tasks with known acceptance criteria: fix a reproducible bug, extract the right figures from a document or complete a reversible workflow in a test account. Keep tools and instructions consistent across models. Record incomplete attempts and manual intervention alongside successful results.
For coding, review the diff and run the relevant checks. For a document question, verify the answer against the underlying page or table. A model can produce fluent prose while missing a qualifier, unit or date. Include those details in the scoring rubric so a polished response does not receive credit for being wrong.
A useful adoption decision for Canadian teams
Compare model charges, tool costs, elapsed time and reviewer effort. Include bilingual material if the application serves English and French users. Preserve the model version and reasoning setting so a later change can be investigated rather than confused with an improvement in the underlying workflow.
The launch makes GPT-6.1 Sol a candidate for a controlled trial. It does not establish that organizations should replace every higher-capability model. The best choice may vary by task: a less expensive system for routine work and another path for difficult cases that fail a defined quality threshold.
Sources and editorial method
We use the linked original statements and reports to distinguish documented facts from our analysis. Company announcements describe their authors’ plans and claims; they are not independent verification of future outcomes. This article is dated and will be revised if material evidence changes. Read our editorial standards and corrections process.
- OpenAI GPT-6.1 Sol release ↗ — Primary release, availability and standard API pricing; checked October 2, 2026. Performance figures are company evaluations.