Menu
Sign In Search Podcasts Charts People & Topics Add Podcast API Blog Pricing
Podcast Image

The Daily AI Show

Gemini 3.1 Pro Preview Jumps Ahead

20 Feb 2026

Transcription

Chapter 1: What is the Gemini 3.1 Pro Preview and its key features?

0.031 - 28.958 Beth Lyons

Hey everybody, welcome to the Daily AI Show. It is Friday. Thank goodness it's Friday. It's February 20th, 2026. I am Beth Lyons. With me in the studio today is Andy Halliday. Back from a snowy adventure. About to go maybe on another snowy adventure. We'll see. We do expect Carl will pop in at some point. Andy, how are you doing today?

0

29.14 - 34.915 Andy Halliday

I'm well, thank you. I'm glad to have an extra day to prep for my trip to Canada next week.

0

35.198 - 45.751 Beth Lyons

That's great. Yeah. And I am not going to Canada next week, which was initially planned. I'll still be in the U.S. for a little while longer.

0

Chapter 2: How does the Google ecosystem integrate AI Studio and NotebookLM?

45.791 - 52.859 Beth Lyons

Big happenings in AI yesterday. Top story, probably the drop of 3.1.

0

Chapter 3: What are the upcoming developments from DeepSeek and GPT-5.3?

55.062 - 60.328 Beth Lyons

Google Pro 3.1. Have you had a chance to read about it, play with it?

0

Chapter 4: How does Arc AGI-III benchmark AI models?

60.308 - 89.32 Andy Halliday

Yeah, I'm sharing a screen here from artificial analysis that shows you how when Google finally drops where they are into the product space, they don't just, you know, kind of eke above Google. the running pack, they jump ahead. So here you see on the artificial analysis intelligence index up here, which is a combination of different factors that illustrate intelligence.

0

89.981 - 93.065 Andy Halliday

And I've isolated the intelligence thing down here.

0

Chapter 5: What new features has Apple Podcasts introduced?

93.125 - 115.881 Andy Halliday

It's interesting that there's a difference this way. But if you look at the combination of its skills, which includes a genetic performance and, you know, other not purely kind of reasoning approaches that this is where, you know, Gemini 3.1 Pro Preview is really shining well above all.

0

116.013 - 150.653 Andy Halliday

anthropic which is right neck and neck with with opus 4.6 max a couple of points ahead of gpt 5.2 thinking high notice how uh you know glm5 and kimmy k 2.5 are right behind the leaders these are both chinese models out there that are available to you DeepSeek V3.2 is pretty far behind on the artificial analysis intelligence index. But now let's look down here at the agentic index.

0

151.714 - 176.086 Andy Halliday

It's interesting to me that Google Gemini 3.1 Pro preview on its agentic skills is way behind Opus 4.6 Max. and regular Opus 4.6 at 64. So that's a pretty big spread back to Gemini 3.1 Pro Preview. What does that mean?

0

Chapter 6: How does WordPress integrate with Claude for website improvement?

176.146 - 193.303 Andy Halliday

For most of us, probably not a lot, unless you're actually using the entire Google ecosystem to set up a harness for multi-agent coding or multi-agent workflow management. You might want to...

0

Chapter 7: What is Post-Visit AI and how does it enhance healthcare?

193.283 - 202.838 Andy Halliday

not focus entirely on Gemini 3.1 Pro Preview. Make Opus available to whatever system you're architecting.

0

203.56 - 224.792 Beth Lyons

Right. It does sort of seem like that was the big question that I saw being asked yesterday. Google 3.1 is significantly better than Google 3.0. The big developer question that I was seeing was, does it still hallucinate in the middle of something?

0

224.772 - 243.473 Beth Lyons

And so having an agentic harness that you trust a little more, like Codex or Opus 4.6 or whatever your agentic harness is that could watch the process as it's going, because it's one of the long thinkers again, right? Yeah.

0

Chapter 8: What is the workflow for show prep and post-show analysis?

243.453 - 262.997 Beth Lyons

Oh, it thought for eight minutes for me or it thought for 15 minutes and then it like generated this very cool thing. In order for that to be successful for you, you either need to be really good at instructing or you have some sort of, I don't know, watcher on the thoughts, maybe like the the.

0

262.977 - 295.21 Andy Halliday

Now, a lot of it depends on memory and the ability to do either recursion, where you process things while creating an intermediate capture of what the results of that are, and then opening up the context window again with new inference based on that new context and so on. So that's this kind of iterative looping process using a Python REPL, storage concept.

0

296.334 - 325.029 Andy Halliday

Let me share, whenever a new model comes out, there's a couple of different benchmarks I go immediately to. One is that artificial analysis index. And the other one is ArcAsia 2.0. And this is the thing that's measuring where these players are in the progression towards an effective accomplishment of artificial general intelligence.

0

325.13 - 357.919 Andy Halliday

And these are really complex reasoning tasks that make it hard for any of the models to really get right. significantly above 60%, 70%, which you see where the players were just recently. Although Gemini 3 DeepThink that came out on February 20th, it's still earlier this month, but it's already at a high expense level. This x-axis down here is cost per AGI...

0

357.899 - 381.249 Andy Halliday

And you can see that GP 5.2 was spending upwards of, you know, $70 per task. And Gemini 3 DeepThink was spending, you know, something in the order of, let's say, $20 a task. Look at this. This is a pretty logarithmic scale looks like here. Yeah. But now look at Gemini 3.1 Pro Preview.

381.289 - 416.025 Andy Halliday

Not only does it jump above Cloud Opus 4.6 and Cloud Sonnet 4.6 on high, which were kind of really in the sweet spot here, generating really high scores on Arc AGI 2, but at a relatively inexpensive rate. Now Gemini 3.1 Pro Preview pushes it back to $1 per task. Yeah. And yet goes upwards of, you know, 75, 78% somewhere in there.

416.687 - 422.803 Andy Halliday

So very impressive performance on Arcade GI2 by Gemini 3.1 Pro, which just came out yesterday.

424.302 - 453.377 Beth Lyons

So I wish that, I love that result. And I agree we're seeing that trajectory. I'm not sure that I have seen something that compares that with consistency. Right. So how many times do you have to answer the question? How many times? Like you got to one shot the question. You one shot at a bunch of questions. It was really impressive. You five shotted these questions.

454.018 - 461.008 Beth Lyons

And therefore, in terms of my looking at where my time is being spent.

Comments

There are no comments yet.

Please log in to write the first comment.