What Is GPT-6.1 Sol? Price and Performance Explained
GPT-6.1 Sol is OpenAI's new model, delivering performance close to GPT-6 Astra's level at one-fifth of the price.
Contents 7
GPT-6.1 Sol is OpenAI's new AI model, released on September 29, 2026. It delivers performance close to GPT-6 Astra's level at one-fifth of the price. For businesses, it stands out in coding, computer-use tasks, and professional workflows.
In short:
- GPT-6.1 Sol pricing: $2/million input, $10/million output tokens; one-fifth of Astra's price.
- It matches Astra's performance on the DeepSWE v1.1 benchmark.
- It scores 7 points higher than GPT-6 Sol on OSWorld 2.0, at less than half the cost.
- Available to Plus, Pro, Business, Enterprise, and Edu users in ChatGPT Work and Codex.
What is GPT-6.1 Sol?
This new model is the latest mid-tier product in OpenAI's GPT-6 family. Its goal is to deliver performance close to GPT-6 Astra's power at a much more affordable price. The company positions this product specifically for coding, computer-use tasks, and professional workflows (openai.com).
As the name suggests, this is an improved version of GPT-6 Sol and Luna, which was announced at the beginning of September. Arriving in less than a week, this update shows OpenAI's rapid iteration strategy.
The API call name has been set as gpt-6.1-sol. Developers can easily update their existing integrations.
How much does GPT-6.1 Sol cost?
Pricing is set at $2 per million input tokens and $10 per million output tokens. By comparison, GPT-6 Astra carries pricing of $10 for input and $50 for output (openai.com).
Input tokens read from the cache cost $0.10/million. Cache write operations are priced at $2.50/million tokens (developers.openai.com).
The model has a massive context window of 1,050,000 tokens. Maximum output is limited to 128,000 tokens. Its knowledge cutoff date is April 30, 2026.
This price advantage matters a great deal for businesses that want Astra-level results but face budget constraints. The new model creates a significant difference in the cost-performance balance.
How did GPT-6.1 Sol perform in benchmark tests?
On the DeepSWE v1.1 software engineering benchmark, it scored at the same level as GPT-6 Astra (openai.com). This means quality very close to Astra's on real-world coding tasks.
On OSWorld 2.0 computer-use tasks, it scored 7 points higher than the previous version. What's more, this was achieved at less than half the cost.
The linear error rate in low-effort mode also improved. The rate dropped from 11.4% to 7.7%, an improvement of roughly 32%.
The average task cost on the Terminal-Bench Science 0.1 test is notable. For the new model, this cost is $5.47. By comparison, Opus 5.5 spends $23.21, and GPT-6 Astra spends $23.80.
| GPT-6.1 Sol | 5,47 |
|---|---|
| Claude Opus 5.5 | 23,21 |
| GPT-6 Astra | 23,80 |
Kaynak: OpenAI
The numbers are clear: the new model runs nearly four times cheaper on similar tasks. This difference provides significant savings for teams with heavy usage.
On the GDP.pdf test, it outperformed Opus 5.5. This was achieved at less than half the cost. On AutomationBench, it delivered similar results at one-third the price, with a 2.2-point difference.
For those who want to compare these results with the price-performance data in our Claude Opus 5.5 article, this offers a separate perspective.
Where can GPT-6.1 Sol be used?
This model is available to Plus, Pro, Business, Enterprise, and Edu users. It can be used in ChatGPT Work and Codex environments.
For developers, the supported effort levels are as follows: low, medium (default), high, xhigh, and maximum. None and minimal levels are not supported on this model (developers.openai.com).
A version called Ultrafast is also coming for Codex. This version promises token generation speeds 8 times faster than standard. It's expected to become available in the near future.
What's the difference between GPT-6.1 Sol and GPT-6 Astra?
GPT-6 Astra, in fact, wasn't released as initially expected. Researchers delayed its launch due to safety concerns (techcrunch.com).
This delay played a role in OpenAI's decision to highlight the new product. The company offered users a model close to Astra's power but more controlled.
In this context, a practical solution emerges in terms of both cost and accessibility. Businesses can get similar-quality results without waiting for Astra.
This kind of model delay is common in the industry. We touched on a similar process in our Gemini 4 Pro leaks article.
Frequently asked questions
What is GPT-6.1 Sol used for?
It's used for coding, computer-use, and professional work tasks. It offers intelligence close to GPT-6 Astra's level at one-fifth of Astra's price.
What's the difference between GPT-6.1 Sol and GPT-6 Sol?
On the DeepSWE v1.1 benchmark, it scored 6.4 points higher than GPT-6 Sol. This was achieved with less effort and lower cost.
Which ChatGPT plans can use it?
GPT-6.1 Sol is available to Plus, Pro, Business, Enterprise, and Edu users. It can be accessed through the ChatGPT Work and Codex interfaces.
Is there a faster version?
Yes, GPT-6.1 Sol Ultrafast is coming soon. This version promises token generation 8 times faster than the standard speed.
GPT-6.1 Sol offers an attractive option for businesses with its price-performance balance. When deciding which model and architecture your software team should move forward with, you can benefit from EngerekTech's enterprise software experience.
Sources
