OpenAI Releases GPT-6.1 Sol and Launches a $500-a-Month ChatGPT Plan

In late September 2026, OpenAI announced two major updates for professional developers and enterprise customers. On one hand, the company unveiled GPT-6.1 Sol, an AI model that delivers flagship-level performance while requiring significantly less compute. At the same time, the developers introduced a new Pro 500 plan priced at $500 a month. The strategy reveals where OpenAI is headed: making basic access to powerful algorithms cheaper while selling maximum generation speed to a premium audience.

An evolution in logic: GPT-6.1 Sol specs

The new GPT-6.1 Sol is a direct evolution of the previous Sol model, not a full replacement for the flagship GPT-6 Astra, as social media users were quick to claim. OpenAI itself describes its performance as “near-Astra performance.” On the DeepSWE v1.1 coding benchmark, the new model did match the flagship and beat the previous Sol version by 6.4 percentage points. But one successful test doesn’t guarantee identical results across every workflow.

The model has already started rolling out in ChatGPT Work and Codex for Pro subscribers. According to the official developer documentation, the algorithm’s specs include:

• a 1.05 million token context window;
• a maximum response length of up to 128,000 tokens;
• support for five reasoning effort levels, from minimal to maximum.

Price is the biggest advantage. On the API, requests cost $2 per million input tokens and $10 per million output tokens. Cached input, meanwhile, dropped to a record-low $0.10 per million tokens — just 5% of the standard price, making the model extremely cost-effective for working with large, frequently repeated databases.

Ultrafast mode for those who won’t wait

Alongside the budget-friendly model, OpenAI introduced the Pro 500 plan. Its main exclusive is Ultrafast mode for the flagship GPT-6 Astra in the Work and Codex interfaces. Subscribers to the regular Plus, Pro $100 and Pro $200 plans don’t get it.

According to Business Insider, OpenAI said at a private presentation that Ultrafast speeds up code generation in Codex eightfold. The company doesn’t make hard guarantees about speed, though: the popular rumor about a fixed “300 tokens per second” isn’t confirmed by OpenAI’s help materials. The mode simply provides minimal latency and server priority, while actual speed will vary dynamically depending on load and the complexity of a given task.

The $500 plan targets large IT teams and programmers first and foremost. For professionals with a heavy daily workload, waiting for the neural network to compile code costs more than a premium subscription.