Today is day four of Tibo's "28 consecutive days of posting or resetting"—something big is coming!
Yesterday, OpenAI announced the release of GPT-6 to all users, including both free and paid subscribers, and today launched the Ultrafast version of the new GPT-6.1 Sol model! Combined with the updated model control system, it enables timely responses to user instructions.
Turbocharged
GPT-6.1 Sol was just released on DevDay on September 29, positioned as "offering capabilities close to Astra at one-fifth the price." The standard API pricing is $2 per million input tokens and $10 per million output tokens, making it one of the most cost-effective leading models available today.
The Ultrafast mode increases the speed by up to 8 times on this basis.
OpenAI refers to it as "Near-Astra intelligence at up to 8x faster speeds than Sol Standard."
This is also OpenAI’s second iteration on the “Ultrafast” concept. Earlier this year, they partnered with Cerebras to release a preview of GPT-5.6 Sol Ultrafast, claiming an output speed of up to 750 tokens/s—14 times faster than standard. While GPT-6.1 Sol Ultrafast’s 8x speed improvement appears more conservative, given that GPT-6.1 Sol itself is significantly more intelligent than GPT-5.6 Sol, achieving an 8x acceleration on this higher capability baseline actually makes it even more impressive.
The significance of this speed for development scenarios is undeniable. OpenAI’s typical use cases include: online troubleshooting, real-time navigation for agents, and any real-time interaction scenario where "every second costs money."
For API pricing, the Ultrafast mode is $12 per million input tokens and $60 per million output tokens, exactly six times the rate of the Standard plan.
OpenAI’s Developer Relations Engineer Dominik Kundel provided a more intuitive comparison: “Intelligence on par with Astra, eight times the speed of Sol, at just 1.2 times the cost of Astra.”
Compared to Astra’s standard API pricing ($10 per million inputs, $50 per million outputs), Ultrafast is only slightly more expensive. Click but it换来 far superior response speeds compared to Astra. If Astra’s speed isn’t sufficient for your use case, Ultrafast offers an option to “pay a little more for a lot more speed.”
The specifications for Ultrafast are enticing, but a crucial piece of information is nearly buried in the announcement.
On Codex and ChatGPT Work, Ultrafast is available only to Pro 500 subscribers, eligible pay-as-you-go Enterprise users, and Edu users.
Pro 500 is OpenAI’s premium subscription plan, launched this year at a monthly fee of $500. This means that users of standard ChatGPT Plus ($20/month) or even Pro ($200/month) must pay extra to access Ultrafast on ChatGPT or Codex, sparking user dissatisfaction.
After testing it, some developers noted that Ultrafast is indeed remarkably fast in XHigh effort mode, but the token consumption rate is equally astonishing—under this mode, the 25x usage quota included in the $500 monthly plan can be "burned through by 1% in just a few minutes." Others questioned the restriction logic itself: "If Ultrafast consumes my usage quota faster, why not let all paying users choose for themselves? I’m willing to accept the trade-off of exhausting my quota more quickly."
The API endpoint is not restricted by subscription tier; developers can directly access the gpt-6.1-sol model and enable Ultrafast mode with pay-as-you-go pricing. This low barrier makes it ideal for teams building agents or real-time applications.
The immediate value of GPT-6.1 Sol Ultrafast lies in its combination of "near-flagship intelligence + extreme speed" at the API level, without requiring you to wait for Astra to be discounted. For everyday users, the improved experience from instant responses may be more tangible—after all, no one enjoys watching a model race headlong in the wrong direction.
What will Day 5 bring? See you tomorrow.
This article is from the WeChat public account "Machine Heart," authored by Machine Heart, edited by Leng Mao.