OpenAI puts GPT-6.1 Sol into overdrive with 8x faster Ultrafast mode

OpenAI has officially launched Ultrafast mode for GPT-6.1 Sol, rolling out across the API, Codex, and ChatGPT Work.

The new tier promises near-Astra level intelligence at up to 8x the speed of Sol Standard. OpenAI is positioning it squarely around critical workflows where every second counts, such as live outage debugging, fast app navigation for agents, and interactive real-time experiences.

Speed comes at an enterprise price

While the performance jump is substantial, getting your hands on it won't be cheap or universally available.

Within Codex and ChatGPT Work, Ultrafast requires a Pro 500 subscription, an eligible usage-based Enterprise plan, or a credit-based Edu account. Even then, Enterprise workspace administrators must manually enable the feature.

For API users, pricing sits at $12 per million input tokens and $60 per million output tokens.

OpenAI also expanded data sovereignty with this release, confirming US and EU data residency support for Ultrafast, alongside newly added EU data residency for GPT-6.1 Sol Fast and GPT-6 Luna Fast.

Community friction and a curious billing catch

Developer reaction in the OpenAI community forums has been a mix of excitement and pushback over the steep paywall.

Multiple users pointed out that without hands-on trials, it is hard to justify jumping straight to a $500 monthly tier just to test if the 8x speed boost changes their day-to-day work. One user paying for a $200 tier noted that while the service tier didn't fail, it didn't deliver noticeable Ultrafast behavior either.

Meanwhile, API tinkerers stumbled upon an intriguing loophole.

One developer reported that after running 50 test calls on Ultrafast with data sharing enabled, their billing dashboard marked the payer as openai, effectively making the developer testing free.

Why it matters

For individual hobbyists or casual coders, Ultrafast simply isn't for you, and upgrading right now makes little financial sense.

However, for engineering teams where minutes of site downtime cost tens of thousands of dollars, or for builders designing sub-second UI agents, an 8x speedup at near-frontier intelligence fundamentally changes system viability.

By locking Ultrafast behind Pro 500 and Enterprise tiers, OpenAI is drawing a distinct line between casual power users and deep-pocketed infrastructure clients. It looks like a classic play to maximize compute margins while testing low-latency infrastructure on paying pros.

The risk, of course, is alienating the mid-tier developer base. If OpenAI doesn't offer limited trials soon, builders on standard plans might look to faster open-weight alternatives before ever experiencing what Ultrafast can actually do.