Omniscope, meet gpt-5

Gpt-5 

OpenAI have released their new flagship model, gpt-5. As well as improved intelligence, it's now unified: no more having to pick different models to control reasoning effort. One model can do it all.

As with gpt-4.1, they've released 3 speed/price tiers, also similarly priced to gpt-4.1. They are:

• gpt-5 - top quality, slower, most expensive
• gpt-5-mini
• gpt-5-nano - lower quality, fastest, cheapest

Head over to the OpenAI models and pricing pages for more details.

Omniscope Report Ninja with gpt-5

Your Omniscope installation, if updated in the past few months, should automatically pick up OpenAI's new models and play well with them. (If not, update to the latest daily build, and let us know.)  I'm assuming you've already tried AI integration and followed the previous guide.

The best place to start is with the "Instant Dashboard", an automatically generated AI dashboard. We recommend the reasoning effort (thinking slider) to "low", otherwise the model spends too long "thinking"; you can do this at point-of-use, and also in Admin by customising default options.

(Note, OpenAI have also introduced a new "minimal" reasoning effort option, which makes it "barely think at all" before responding. If you really want a quick reply, we'll be supporting this soon. They've also added a "verbosity" setting, and we will expose that as an option, subject to demand.)

Here are the 3 tiers and an example result, using a dataset of aid worker incidents. 

gpt-5-nano:

gp5-mini:

gpt-5:

At time of writing, it appears gpt-5 is suffering from over demand, and performance is less than it should be, and fluctuating dramatically. So, "thought for 22 seconds" is somewhat more than I'd expect, for "low" reasoning effort. But all 3 models are quite capable of understanding the data and dashboard concepts, and generating a starter dashboard.

Omniscope AI Completion block with gpt-5

Let's jump from Report Ninja over to the Omniscope workflow and look at how we can use gpt-5 in your data prep. Here's an example using gpt-5-mini to generate descriptions of the hurricanes in our 2003-2005 hurricanes demo dataset. Again, I'm using reasoning effort "low", because I'm impatient.

Now, this block performs an fresh AI request for each and every input record. In my case I've got 55 hurricanes from 2003-2005, so after aggregating by hurricane, I've sorted by wind speed and kept only the first 10. See earlier comments about being impatient, and about gpt-5 being in high demand!

Now how about an Instant Dashboard?

(with some minor adjustments)

That's all for now. This space moves fast, so it won't be long until the next update. Meanwhile, please contact us if you have any feedback or questions.