“A YAWA RA.”
That was literally the first thing I said early in the morning as I routinely browsed through my daily tech news feed, still half-awake, and then I saw this BULLSHIT blog from OpenAI about their newest flagship model, GPT-5.6 Sol.
And apparently, it’s even higher in performance than the goddamn Anthropic MYTHOS????
Ur telling me we’re currently in a nuclear bomb vs. nuclear bomb type shi?
I stand my ground with my statement when I said that this is still a net positive for humanity, because having two powerhouses in AI development is better than having one monopoly controlling the whole thing. But damn, the amount of one-ups they keep doing to each other makes every other competition look like a side quest at this point.
Y’all can check out OpenAI’s blog directly through here:
https://openai.com/index/previewing-gpt-5-6-sol/
But as a TL;DR, Sol is apparently the most capable out of the new classification of AI models they’re releasing, along with Terra and Luna, which is essentially their own version of Anthropic’s Opus, Sonnet, and Haiku, respectively.
However, their Sol pretty much outperforms Fable’s and, by extension, Mythos’ performance across a broad range of fields, including cybersecurity, agentic coding, biology for some reason, and other areas that honestly sound like the tech industry is just casually speedrunning the future now.
It’s also currently under review under the U.S. cybersecurity lockdown to make sure everything is safe and properly gatekept, similar to what they’re doing to Mythos under the Glasswing Project.
That’s cool and all, but the biggest “fuck you, Claude” moment to me by far is the pricing.
OpenAI is clearly leaning into the affordable choice profile between the two supergiant AI firms. We already saw that with the previous release of GPT-5.5, back when Anthropic’s session limits were insanely ahh-bysmal.
Now, they’re starting at $30 per 1 million output tokens, which is an insanely cheaper alternative compared to Fable’s debut pricing at $50 per 1 million output tokens. It’s much closer to Opus, which was around $25, except this is for allegedly insane flagship-level performance.
Currently, it’s only available to a select few trusted organizations in a soft-opening setup for cybersecurity purposes. But before this, they definitely did their due diligence in protecting their models.
Red teaming. Safety checks. And apparently using 400 hours of over a bajillion motherfuckin datacenter processing powers to basically pen-test their own models.
Insane to think that models are now improving themselves. Like, can’t this be considered AGI at this point? Damn.
Anyways, when this releases, hell breaks loose.
But at this point, we’ve been through so many “it’s over” moments that honestly, I’m desensitized now. I guess I’m just tired of complaining and just wanted to lock in on the development of my projects instead of fear mongering.
It is what it is. SWE jobs were already cooked and grilled way before this dropped. They’re pretty much just kicking over a corpse, man.
So we keep vibin my dudes, I suppose.
