Ben (no treats) argues massive models are overrated for daily coding due to latency, preferring a faster GPT-5.5

That reads like a solid initial assessment. GPT-5.6 will likely offer a better price-performance ratio than Fable 5; however, given the recent announcement that Fable 5 already has a newer version (5.1?), it seems logical that Fable will likely remain the better overall model for the time being.

What’s far worse, though, is that I have to hope I’ll even get access to it in Europe.

@kimmonismus Exactly 💯

@kimmonismus Doubtful Fable 5.1 would be as good as the Fable we saw considering there's a reason it's restricted and releasing a bigger/better Fable would likely result in the same outcome.

@kimmonismus the labs need to settle this down, ai needs to be decentralised else the world is going to be doomed :(

@PradyuPrasad 😭

@tenobrus > i have no access

haha pleb

In my opinion, I hope they bring more censored versions of these models to the global market and just block certain prompts—you know, the ones involving malicious use. Otherwise, this just isn't going to work. If companies only sell their best AI to a hundred other companies, they'll go bankrupt; they absolutely have to open it up globally. The Trump administration is very disorganized; they can't seem to get things done properly.

@kimmonismus that's a fair point on the perf tradeoffs tbh

@kimmonismus optimistic future can't be achieved this way....we can just hope for the best, it shouldn't go in that route I believe and pray

@kimmonismus The worst is that we are suffering the consequences of cooling data centers for tech we don’t get to enjoy (even if we can pay for it)

@kimmonismus the versioning war is already starting and the benchmarks havent even settled yet

@tenobrus so apparently new major version number = new pretrain and new minor version number = new post-train on the same pretrain?

@arayush01 I think we are past this point, sadly

@tenobrus version number + timing + lack of price change = the pretrain is clearly the same imo

theyll have a big boy train come fall tho

@kimmonismus Worth noting Fable 5 isn't region-gated right now, it's suspended globally under a US export-control order since June 12. Nobody has access, not just Europe

@tenobrus Based off the initial benchmarks and rumors at face value what I think we can conclude their post training sauce is still the best though right?

Again that’s just shooting the shit off narrow performance they showed, but they don’t game these usually

@kimmonismus We have to thank the incompetent European regulators for that. And I don’t see them changing course. Lagarde and her technocrats are taking all the wrong decisions and national politicians still seem to believe they can beat the odds on their own 🤷‍♂️🤦‍♀️🤦‍♂️🤦🏽

@kimmonismus cost per token means nothing if your code doesn't work.

Fable's more reliable on code tasks. that's what saves you actual time.

@kimmonismus Just shut up about Fable. Fable is a good model and that’s it. For pure math and open math problems it is as good or worse than gpt 5.5 pro.

@kimmonismus So multiple Sol agents on max effort can beat Fable. What happens If you Run multiple Fable agents on max effort?