Latest / Ctrl AI Profit / Ep. 150 | OpenAI Just Gave You Three Different Brains to Choose From
Transcript
- 0:20SPEAKER_01: Frank, I feel like every time I get comfortable with one version of ChatGPT, they announce another one.
- 0:28SPEAKER_00: Now, there are three.
- 0:30SPEAKER_00: Three in one family, yes.
- 0:31SPEAKER_00: OpenAI recently launched a preview of what they're calling GPT 5.6.
- 0:37SPEAKER_00: And instead of one model, they released three: Saul, Terra, and Luna.
- 0:42SPEAKER_00: Different price points, different capabilities, different intended uses.
- 0:46SPEAKER_01: Okay, walk me through them because I've already confused my clients enough.
- 0:50SPEAKER_01: Talking about AI models.
- 0:52SPEAKER_00: Think of it this way: Saul is the flagship, maximum reasoning, multi-step problem solving, what they call ultra modes.
- 1:00SPEAKER_00: This is the most powerful version they've built.
- 1:03SPEAKER_00: Terra is the everyday model, roughly the same quality as what you've been using, but cheaper to run.
- 1:09SPEAKER_00: And Luna is optimized for speed and volume.
- 1:12SPEAKER_00: High output, low cost, good for things like generating a lot of drafts fast or handling customer interactions at scale.
- 1:21SPEAKER_01: So they've basically segmented the product the same way any smart company does.
- 1:26SPEAKER_01: You've got your premium, your standard, and your high volume tier.
- 1:32SPEAKER_00: Exactly.
- 1:33SPEAKER_00: And the pricing follows that logic.
- 1:35SPEAKER_00: Terra and Luna are significantly cheaper than Saul.
- 1:38SPEAKER_00: The idea is you pick the right brain for the right job instead of using the most expensive model for everything.
- 1:44SPEAKER_01: Which honestly makes sense.
- 1:46SPEAKER_01: I don't use my best employee to stuff envelopes.
- 1:50SPEAKER_00: That's a perfect analogy.
- 1:52SPEAKER_00: Saul is your senior strategist.
- 1:54SPEAKER_00: Terra is your capable generalist.
- 1:56SPEAKER_00: Luna is your high volume operator.
- 1:59SPEAKER_00: You use them appropriately.
- 2:01SPEAKER_01: Now, what's the catch?
- 2:03SPEAKER_01: Because I noticed you said this is a preview.
- 2:06SPEAKER_01: Limited preview.
- 2:07SPEAKER_00: And here's the unusual part.
- 2:09SPEAKER_00: OpenAI did this at the request of the US government.
- 2:13SPEAKER_00: They're doing staged rollout with safety validation and red teaming before general availability.
- 2:18SPEAKER_00: Trusted partners get access first.
- 2:21SPEAKER_00: General availability is weeks out, not months.
- 2:24SPEAKER_01: The government asked them to slow down the release of a new AI model.
- 2:29SPEAKER_00: Specifically asked for a controlled rollout with safety checks before the public gets it, and OpenAI agreed.
- 2:36SPEAKER_00: That's a significant shift from how AI releases have worked up until recently.
- 2:41SPEAKER_00: Why does that matter to a business owner?
- 2:44SPEAKER_00: A few reasons.
- 2:45SPEAKER_00: One, it signals that these models are powerful enough that even the government is paying attention.
- 2:51SPEAKER_00: Two, it means when you do get access, this version has been vetted more carefully than prior releases.
- 2:57SPEAKER_00: Three, it's a preview of how AI deployment is going to work going forward.
- 3:01SPEAKER_00: More oversight, more staged rollouts, less Wild West.
- 3:05SPEAKER_00: The Wild West part was kind of fun though.
- 3:08SPEAKER_00: It was also unpredictable.
- 3:10SPEAKER_00: Businesses would build workflows around a model, and three months later, that model would change or disappear.
- 3:17SPEAKER_00: The more regulated cadence actually helps with planning.
- 3:20SPEAKER_00: Fair point.
- 3:21SPEAKER_01: Okay, so for someone running a small business, how do they think about which of these three models to actually use?
- 3:29SPEAKER_00: Start with the use case.
- 3:31SPEAKER_00: If you're generating marketing copy, writing emails, summarizing documents, answering customer questions, Terra is probably your default.
- 3:41SPEAKER_00: It's cost effective and capable enough for most tasks.
- 3:44SPEAKER_00: If you're doing complex analysis, strategic planning, multi-step research, that's Sol territory.
- 3:52SPEAKER_00: If you're running high volume, like generating hundreds of product descriptions, or handling large customer service cues, Luna.
- 4:00SPEAKER_01: And the pricing difference between them is significant enough to matter?
- 4:05SPEAKER_00: For most small businesses doing light to moderate usage, the difference is a few dollars a month.
- 4:10SPEAKER_00: But if you're integrating this into a workflow that runs thousands of requests, which is where the real automation value is, the difference between SOL pricing and Luna pricing could be 10 to 1.
- 4:22SPEAKER_01: Which is why it matters to understand what you're actually using before you build the automation.
- 4:28SPEAKER_00: Right.
- 4:29SPEAKER_00: The expensive mistake is using your premium model for tasks that don't need it.
- 4:33SPEAKER_00: It's like running your diesel truck on premium gasoline.
- 4:36SPEAKER_00: It won't hurt anything, but you're wasting money.
- 4:39SPEAKER_01: Let me ask the practical question.
- 4:41SPEAKER_01: Right now, can my business actually use these?
- 4:45SPEAKER_00: The preview is limited to trusted partners, so probably not today, but by the time general availability hits, which OpenAI says is weeks away, these will be accessible through the standard API and through ChatGPT subscription tiers.
- 4:58SPEAKER_00: You don't have to do anything special.
- 5:00SPEAKER_00: They'll show up in your existing tools.
- 5:02SPEAKER_01: What I keep coming back to is that OpenAI is now basically operating like a product company with a product lineup.
- 5:11SPEAKER_01: You've got different SKUs for different customers.
- 5:15SPEAKER_00: And that's exactly what enterprise customers have been asking for.
- 5:18SPEAKER_00: Not one size fits all, but a menu of options with clear price-to-performance trade-offs.
- 5:24SPEAKER_00: A hospital system has different needs than a food truck.
- 5:27SPEAKER_00: Now they can both be open AI customers without the hospital paying food truck prices or the food truck paying hospital prices.
- 5:34SPEAKER_01: Which is how every other enterprise software company has operated for decades.
- 5:39SPEAKER_01: AI is finally catching up to normal business logic.
- 5:44SPEAKER_00: The maturing of the industry in real time.
- 6:15SPEAKER_00: The future just got a little more organized.
- 6:19SPEAKER_00: Three brains, three prices, one company.
- 6:22SPEAKER_00: Welcome to Enterprise AI.
- 6:25SPEAKER_00: Let's talk about what Saul actually unlocks.
- 6:28SPEAKER_01: You said subagents.
- 6:30SPEAKER_01: What does that mean in practice?
- 6:33SPEAKER_00: Subagents means the model can spin up smaller AI tasks to complete parts of a bigger job.
- 6:38SPEAKER_00: So if you ask Sol to research a competitor and write a competitive analysis, it doesn't just do that linearly.
- 6:44SPEAKER_00: It can farm out the research piece, the writing piece, the formatting piece, running them in parallel, and then combine the results.
- 6:53SPEAKER_01: That's not one AI, that's a team of AIs working together.
- 6:58SPEAKER_00: That's exactly what it is.
- 7:00SPEAKER_00: And that's why Sol is priced higher.
- 7:02SPEAKER_00: It's not just a smarter chat window, it's a coordinator that can manage other AI processes.
- 7:08SPEAKER_00: For a small business, when would you actually need that?
- 7:12SPEAKER_00: The use cases are complex, ongoing workflows, monthly financial summaries that pull from multiple sources and write the narrative.
- 7:20SPEAKER_00: Marketing campaigns that research trends, generate copy, and analyze performance all in one job.
- 7:25SPEAKER_00: Client reporting that aggregates data and formats a professional document.
- 7:30SPEAKER_01: Things that right now take a human a few hours to pull together.
- 7:34SPEAKER_00: Things that could run overnight and be waiting for you in the morning.
- 7:38SPEAKER_00: That's the practical version of what subagents enable.
- 7:41SPEAKER_01: That honestly sounds like having a junior analyst on staff without the salary.
- 7:47SPEAKER_00: That's the trajectory, which is why getting familiar with how these tiers work and what each one is actually capable of puts you ahead of the business owners who are still treating AI like a search engine.
- 7:59SPEAKER_00: Let me ask one more practical question.
- 8:02SPEAKER_01: Right now I'm paying for ChatGPT Plus.
- 8:05SPEAKER_01: Does any of this change what I'm paying?
- 8:07SPEAKER_00: Not immediately.
- 8:09SPEAKER_00: The three-tier rollout is happening on the API side first, then filtering into subscription tiers.
- 8:14SPEAKER_00: What you'll eventually see is clearer labeling inside ChatGPT for which model is handling your request, and likely pricing options to upgrade to Seoul or optimize costs with Luna, depending on what you're doing.
- 8:27SPEAKER_01: So this is also OpenAI finally giving regular users more control over the cost to capability trade-off.
- 8:36SPEAKER_00: Exactly.
- 8:37SPEAKER_00: Right now you pay one price and you get one model.
- 8:40SPEAKER_00: The new structure lets you pay for more when you need it and pay less when you don't.
- 8:45SPEAKER_01: That's just smarter software pricing.
- 8:48SPEAKER_00: It's been standard in every other enterprise software category for 20 years.
- 8:53SPEAKER_00: AI is catching up.
- 8:55SPEAKER_01: Thanks for listening to Control AI Profit.
- 8:58SPEAKER_01: Forward this to someone trying to figure out which AI model to use.
- 9:02SPEAKER_01: It'll save them an hour of Googling.
- 9:05SPEAKER_01: See you tomorrow.