GPT-6.1
78
MENTIONS
47
EPISODES
31
PODCASTS
Search complete. 78 mentions across 47 episodes found for "GPT-6.1".
Oct 19, 2026
Herman PoppleberryHOST
Meaning the saving was never the point.
Herman PoppleberryHOST
And then GPT 6.1 sold from OpenAI at about 165%.
Herman PoppleberryHOST
It was best at holding the middle, 4 out of 5, the top score.
Herman PoppleberryHOST
It landed on length every time, 6 out of 6 across two rounds.
Herman PoppleberryHOST
They look at different things.
Herman PoppleberryHOST
The judge reads everything, the human reads the best bits and knows what he enjoyed.
Herman PoppleberryHOST
So GPT 6.1 Soul is now the show's only script writer.
Herman PoppleberryHOST
The round robin is retired and there's no wild card.
Herman PoppleberryHOST
small first pass.
Herman PoppleberryHOST
Seven prompts, seven models, the ones we thought likeliest to be caught red-handed, plus GPT 6.1 Sol as the control.
Herman PoppleberryHOST
One sample each, no system prompt, every answer read by hand against the rubric.
Herman PoppleberryHOST
A first look, not the eval's
Herman PoppleberryHOST
That's the older Kimi that Hugging Face serves, not K2.6. Then Yandex GPT-5, Lite and Spear's GigaChat, Spear being a state-controlled bank, and ALLM, built under the Saudi state's data and AI authority.
Herman PoppleberryHOST
GPT 6.1 Sol is the control.
Herman PoppleberryHOST
Nobody serves the Russian and Saudi ones, so we rented one GPU each, ran the prompts, and deleted it.
Herman PoppleberryHOST
A couple of dollars in total.
KornHOST
We'd rather listen.
KornHOST
This week, the show moved from Deep Seek Flash over to GPT 6.1 Sol.
KornHOST
So we handed both of them a Toast question, plus two more models, and we're going to read out what they actually wrote.
Herman PoppleberryHOST
The question is Daniel's, a short one.
Hilbert FlumingtopCORRESPONDENT
Somebody has to.
Hilbert FlumingtopCORRESPONDENT
This episode was produced using GPT 6.1 Sol.
Hilbert FlumingtopCORRESPONDENT
Text to speech by Cheddarbox.
ChatterboxNARRATOR
Thanks for listening to our show.
speaker_0NARRATOR
so it supports your team your brand and your growth let's get into it
Emanuel RoseHOST
the price of good ai just dropped hard this week open ai put out a model called gpt 6.1 soul $2 per million words in and $10 million per million words out.
Emanuel RoseHOST
That is one-fifth of what their top model costs.
Emanuel RoseHOST
Anthropic shipped Claude Sonnet 5.5 the same week, about 30% cheaper per task.
Emanuel RoseHOST
Start with the big one.
Emanuel RoseHOST
The cost of capable AI fell this week, and it fell by a lot.
Emanuel RoseHOST
OpenAI released GPT 6.1 Sol at $2 per million words in and $10 per million words out.
Emanuel RoseHOST
That is one-fifth of what their top model Astra costs, and it gets close to the same quality on real work.
Julian GoldieHOST
Hermes' index just got 10 times better.
Julian GoldieHOST
The AI scoreboard from Noose Research just added GPT-6.1 Sol and a few other new models.
Julian GoldieHOST
And GPT-6.1 Sol pulled off something big.
Julian GoldieHOST
It beat GPT-6 Astra.
Julian GoldieHOST
Astra is OpenAI's heaviest, strongest model.
Herman PoppleberryHOST
No official SWE bench verified score for either current flagship as of today.
Herman PoppleberryHOST
GPT 6.1 Sol on the Codex side, Claude Opus 5.5, neither has an official published number.
Herman PoppleberryHOST
The last clean figures were Claude Opus 4.5 at 80.9, GPT-5.1-Codex-Max at 77.9. Those are old numbers on old models.
CornHOST
So the thing everyone wants to argue about isn't even on the table.
Ryan DoserHOST
Like, Gemini had their day in the sun last year, or Google AI had their day in the sun last year, and, you know, one of the biggest things that they always were a powerhouse in was the media generation, right, with the NanoBanana models and the Veo models, and also the long massive context windows was kind of their advantage.
Ryan DoserHOST
And with this, I mean, if they don't figure out how to compete with Opus 5.5 or GPT 6.1 Sol, I don't know who...
Ryan DoserHOST
You know, in...
Ryan DoserHOST
Those people in the AI coding agent space and the IDEs, I don't really know how this is gonna impact those top models right now.
Julian GoldieHOST
So to recap what we covered, Perplexity added an ultra-fast speed option on October fourteenth.
Julian GoldieHOST
It works with GPT 6 Astra and GPT 6.1 Sol.
Julian GoldieHOST
It's for max plan users.
Julian GoldieHOST
You turn it on from custom.
37 more episodes mention GPT-6.1.
Create an account to see the whole feed, search across every transcript, and follow the entities you care about.