When Every Second Costs You a Customer
Picture this: a customer messages you on WhatsApp at 11pm asking about a product, and your AI support agent takes 20 seconds to produce a so-so response. Or you’re generating a monthly report, and you have to wait a full coffee break for the AI to pull everything together. Speed, for a small team, is not just a convenience—it’s often the difference between winning a sale and watching the customer drift to a competitor.
That’s why a quiet update from OpenAI on Thursday should get your attention. The company rolled out Ultrafast, a new mode that makes its most powerful model, GPT-5.6 Sol, work at 14x the speed of standard processing—delivering up to 750 output tokens per second. But before you go rewriting your entire tech stack, let’s break down what this actually means for a Malaysian SME like yours.
TL;DR: OpenAI’s Ultrafast mode makes GPT-5.6 Sol run 14x faster, outputting up to 750 tokens per second, powered by a Cerebras partnership. It’s currently in preview for select customers, with broader access coming as capacity grows. For you, the practical win is faster customer responses, quicker data analysis, and more work squeezed into the same lunch break—but you don’t need to jump on it today.
What “Ultrafast” Actually Means
Let’s set the jargon aside. When an AI model writes a response, it generates “tokens”—small chunks of text, roughly a syllable or word for English, a bit messier for Bahasa Malaysia. A speed of 750 tokens per second means that in the time it takes you to blink twice, the model can produce a short paragraph. That’s dramatic compared to the standard version, which moves at roughly one-fourteenth of that speed.
| Metric | Standard GPT-5.6 Sol | Ultrafast Mode |
|---|---|---|
| Output speed (tokens per second) | ~54 (implied) | up to 750 |
| Speed multiplier | 1x | 14x |
| Typical response time for a 100-token email | ~1.9 seconds | ~0.13 seconds |
| Availability | General | Preview for select customers |
Source: TechCrunch and derived calculations
OpenAI says this kind of speed was previously only possible with smaller, specialised models, but Ultrafast points to “more useful work per second” from a full-scale model. The company also notes that competitors like Anthropic have launched faster modes, but Claude’s fast mode doesn’t match this speed. Ultrafast is powered by a partnership with chipmaker Cerebras, and it’s currently only available to a small group of customers—though OpenAI says access will expand over time.
For a business owner, the headline number is not the token count. It’s what that speed unlocks in your daily operations: real-time conversation, instant data summarisation, and no more waiting around for the AI to finish its homework.
How This Applies to Malaysian SMEs
Here’s where it gets practical. The most obvious use case is customer service. If you run an online store on Shopee, Lazada, or your own WooCommerce site, your support team is racing against the clock. When a customer asks about delivery to Sabah or your return policy, a 10-second delay doesn’t feel like much—but multiplied across 30 chats a day, it adds up. With Ultrafast, an AI assistant can draft instant, localised responses in Manglish or Bahasa Malaysia without making a customer feel like they’re talking to a machine on a slow connection. For a team of 3 support staff, this effectively gives you an extra pair of hands.
Another area is reporting and analysis. Many Malaysian SMEs depend on monthly sales reports from Google Analytics, Shopee seller centre, or accounting software. Normally, you’d export the data, ask ChatGPT to find trends, and wait a minute or two. Ultrafast cuts that down to seconds, which means you can run a quick check in the middle of a conversation with a potential investor or your business partner, instead of saying “let me get back to you.” It makes ad-hoc decision-making much more fluid.
There’s also content production for marketing. If you’re a small F&B brand trying to keep up weekly Instagram posts and WhatsApp broadcasts, you might be spending hours writing captions. With a faster model, you can batch-generate 14 times as many drafts in the same time, then spend your effort on editing the good ones. That’s not about working harder—it’s about reducing the time your team spends waiting for output.
But here’s the honest caveat: Ultrafast is not available to you yet. It’s in a closed preview, and OpenAI is deliberately expanding capacity slowly. That doesn’t mean you should ignore it. Instead, it gives you a clear signal about where the industry is heading—and a reason to start preparing your workflows so that when faster models arrive, you can plug them in without a scramble.
Speed is not just about getting answers quicker. It’s about changing what you can do—like answering 20 chats while your competitor is still waiting for the first reply to finish generating.
Practical Takeaways
- Don’t rebuild your systems today, but do map out which of your processes are slow and why. Is it the model, the internet, or the human review step?
- Test AI faster modes on your current setup once they become available. Even a 2x speed boost can make a measurable difference in response times.
- Start with a single use case—customer service or report generation—and measure the time saved before expanding.
- Keep an eye on local AI providers and platforms like AutoRunBiz that may integrate fast models into their automation tools, so you don’t have to do the technical work yourself.
- Remember that speed without accuracy is useless. Always validate critical outputs, especially for financial figures or legal statements.
The Bigger Picture
For years, the trade-off in AI has been between power and speed. Small models were quick but limited; big models were smart but slow. Ultrafast suggests that gap is closing. With chipmakers like Cerebras pushing the hardware, and OpenAI pushing the software, “real-time” will soon be the default for any AI tool you use—not a premium feature.
For Malaysian SMEs, this matters because you already run lean. You don’t have large teams to absorb inefficiencies. When AI can produce work 14x faster, the barrier to automating a task gets much lower. That 3-hour report becomes a 3-minute check. That backlog of customer queries that you only handled twice a day becomes a live conversation.
The long-term impact is not that you’ll spend less time waiting. It’s that you’ll question why you ever accepted that kind of lag in the first place. For a small business, time is not just money—it’s the resource you can never buy more of. And anything that gives you more of it, even if it’s just a preview today, is worth watching closely.
Ready to Streamline Your Operations?
Your business should run itself. AutoRunBiz deploys AI agents to automate your daily operations — WhatsApp orders, invoicing, customer follow-ups, and accounting. Book a free 15-min ops audit to see where automation fits your business →