GPT-6 Sol and the AI price war

SPEAKER_1 (0:00)

Your heart can tell you a lot about your health. Apple Watch Series 12 measures your heart rate every five seconds with the most accurate heart rate sensing in a wearable. So your vitals app, now with heart rate variability, can tell you when something is off. And your readiness score can let you know when to rest and when to push. Hear the story in every heartbeat with Apple Watch Series 12

The features described are for wellness purposes only and not for medical use. iPhone 11 are later required based on Apple conducted study of heart rate accuracy August 2026 Visit apple.com/applewatchseries12.

SPEAKER_2 (0:30)

This episode is brought to you by PayPal. You know how a mom's bag has everything? Sunscreen, snacks, a stapler?

The new PayPal app is like that, but for your money. Shop, pay, manage your account and earn rewards all in one place. And with purchase protection on eligible items, biometric security and pass keys, you're protected at every step. Download the new PayPal app to get started. See paypal.com/protectionterms.

SPEAKER_3 (0:59)

When you need to build up your team to handle the growing chaos at work, use Indeed Sponsored Jobs. It gives your job post the boost it needs to be seen and helps reach people with the right skills, certifications and more. Spend less time searching and more time actually interviewing candidates who check all your boxes. Listeners of this show will get a $75 sponsored job credit at indeed.com/podcast.

Terms and conditions apply. Need a hiring hero? This is a job for Indeed Sponsored Jobs.

SPEAKER_4 (1:29)

OpenAI just released GPT-6 Sol and GPT-6 Luna, basically cutting their own previous generations pricing by half while pushing capability scores up across the board.

SPEAKER_5 (1:41)

Yeah. That release dropped while the industry was still trying to process Anthropic launching Claude Opus 5.5. You have these two massive engineering organizations pushing production-ready architecture into the public sphere, practically overlapping each other.

SPEAKER_4 (1:55)

Right. Which leads us with this core question of how business strategies and daily workflows actually function when the cost of executing complex multi-step reasoning drops to literal cents.

SPEAKER_5 (2:08)

Because if you want to understand the economic reality of that, you really have to look at the new price floor they're establishing here.

SPEAKER_4 (2:14)

Yeah. Sol is sitting at $2 per million input tokens and $10 for output.

SPEAKER_5 (2:19)

Right.

SPEAKER_4 (2:19)

And Luna is even cheaper, sitting at just $0.10 for input and $0.50 for output. And both of them maintain that $1.05 million token context window and a $128,000 maximum output threshold.

SPEAKER_5 (2:33)

You look at Luna specifically and there is this very intentional devaluation of compute happening. I mean, Luna previously took an 80% price reduction, dropping from $6 down to that $0.50 per million output tokens within a really narrow window.

SPEAKER_4 (2:46)

Which, I mean, looks like a clear win for users.

SPEAKER_5 (2:49)

It is, but a price rise to the bottom isn't automatically a sign of the underlying silicon getting cheaper to run. It functions as a defense mechanism, really. The market is absorbing this flood of highly competent, open-weight Chinese models, and holding on to enterprise market share just means you have to starve those new competitors of margin.

SPEAKER_4 (3:07)

Yeah, the defensive posture definitely works, but the performance jump on the OpenAI side is still highly visible in the data Wade Foster shared from Zapier. They ran GPT-6 Sol through Automation Bench, which is 657 of their toughest workflows, and it scored 33.2% on high effort tasks, beating the 28.77% score of the older GPT-56 Sol.

SPEAKER_5 (3:31)

And doing all of that for half the cost, but the specific customer support example from those tests is what really clarifies the distinction between those two percentages. They task the models with turning incoming conversations into tickets, and then updating customer profiles based on strict company policy.

SPEAKER_4 (3:47)

Right, and the older 5.6 model basically just identified the issue and labeled the conversations, but left the profiles totally untouched. Whereas the new GPT-6 Sol successfully completed both the labeling and the profile update.

SPEAKER_5 (4:01)

If you're not subscribed yet, take a second and hit follow on whatever podcast app you're using. It helps us keep making this.

We appreciate you being here.

SPEAKER_4 (4:10)

Because identifying a problem and actually executing a multi-step solution are entirely different computational tasks. We're crossing this threshold where the AI transitions from an assistant that highlights information for you to an active participant that finishes the job entirely.

SPEAKER_5 (4:26)

Exactly. And that directly dictates how they're splitting capabilities across their tier system now.

SPEAKER_4 (4:31)

Yeah, Astra remains the high capability flagship model.

Sol is positioned for complex coding and agentic workflows. And then Luna is engineered for high-volume repetitive tasks like document summarization and raw information extraction.