#

Key takeaways

• OpenAI introduced GPT-6.1 Sol as an upgrade to GPT-6 Sol.

• The company says it improves coding, workflows, computer interaction, and factual accuracy.

• OpenAI reports strong results on some evaluations, but not universal replacement of Astra.

• The model is available through the API and in several ChatGPT Work and Codex plans.

• A faster Ultrafast version is planned for the coming days.

OpenAI introduces GPT-6.1 Sol

OpenAI introduced GPT-6.1 Sol on September 29, 2026. The company presents it as an upgrade to GPT-6 Sol. It says the model improves coding, professional workflows, computer interaction, scientific work, and factual accuracy.

The announcement relies on company-reported evaluations. That matters because the results are not independent guarantees for every task. OpenAI also notes that some tests are deliberately challenging. Their failure rates should not be treated as ordinary-use rates.

What OpenAI says improved

OpenAI says GPT-6.1 Sol performs better on several practical tasks. These include coding, professional document work, and automated business workflows. The company also says the model shows progress in scientific work and factual accuracy.

On DeepSWE 1.1, OpenAI says the new model matches GPT-6 Astra while costing approximately one fifth as much. That is a specific comparison from the company. It suggests a focus on efficiency as well as capability.

Where Astra still leads

OpenAI does not present Sol as a universal replacement for Astra. In its scientific evaluation, Astra still leads on the hardest tasks. That means the announcement supports improvement, not total dominance across every benchmark.

This distinction is important for readers. A model can perform well in one area and still trail another model in a different test. OpenAI's own wording keeps that boundary clear.

Availability and pricing

OpenAI says GPT-6.1 Sol is available through its API under the identifier `gpt-6.1-sol`. It also lists availability in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise, and Edu users. The company says it is not yet available in Chat.

OpenAI lists standard API pricing at two dollars per million input tokens, ten cents per million cached input tokens, and ten dollars per million output tokens. These are the prices stated in the announcement. Readers should use them as the source's published figures.

Safety and limitations

OpenAI says safety evaluation results show improvements in respecting restrictions and in transparency about limitations. That is a positive signal, but it comes from the company's own testing. The announcement does not claim that all safety concerns are removed.

The source also warns against overreading the failure rates from these tests. They are designed to be difficult. So the numbers should not be used as a simple proxy for everyday performance.

What the announcement does and does not prove

The release supports a narrow conclusion. OpenAI is saying GPT-6.1 Sol is stronger in several areas and more efficient in at least one benchmark comparison. It is also saying the model is now available in specific products and via the API.

The release does not prove universal superiority. It does not establish that Sol replaces Astra in every setting. It also does not provide independent validation of the company's comparisons.

Morocco relevance

The source reports no Morocco-specific rollout, customer agreement, or regulatory approval. For readers in Morocco, the global lesson is to separate vendor claims from independent validation before planning adoption.

Bottom line

GPT-6.1 Sol is positioned as a practical upgrade with broader workflow value. OpenAI emphasizes coding, document work, scientific tasks, and factual accuracy. It also highlights availability across API and selected product tiers.

At the same time, the announcement stays careful. It presents company-reported benchmarks, notes remaining limits, and avoids claiming universal replacement of Astra. That makes the release useful, but not definitive, for judging real-world performance.