AdSponsored0:00
Continue watching?
Resume from 00:00
00:00 / 00:00
Sleep timer ended
⏪ 10s
⏩ 10s
1M views
1d ago
Mux is the video API that translates audio and answers questions about your video. Get $50 in free credits: https://mux.com/fireship Anthropic just dropped Fabl…
...more
98 Comments
If an AI on the horse benchmark is getting near a 10/10 score, we should have a cat benchmark. Cats are harder to model and animate I think.
"These Chinese models are excellent" 🙃<br>Fukunggg hilarious
typescript, lol
Yup, fable 5.2 diagnosed my 4 years health issue in one shot and suggested the best treatment option after seeing dermatologists, gps, surgeons; they had all been inconclusive. I gave it a huge dataset and full context, it interpreted all the files and one-shotted everything. Truly amazing yet terrifying; it shows just how good it is.. There is no "Bubble".
You seem like you've invested in Anthropic. You do realize it's not even worth $100 Billion, right ?
What about Gemini Carbon though?
As a software dev, these last years have been such torture. Career switch impending
It is warping the mesh, though.That's how rigging works.
I didn't think I would say this, but Typescript is impressive.
Has any humanoid submitted a hand-rolled horse yet to see how they compare?
Aw sweet, an update for Game Dev Story. Let's see what's new. Alright, I can have 1000 agents whip me up an RPG, Schmup, Sandbox, Sports, or Fighting game.
wtf is up with the back legs of the horse, its like there is 2 legs rendered on top of each other for each back leg ?
Jensen Wong 😂😂😂
at this point we must ask ourselves if it's possible to double the bones to horse ratio by 2080
Using that domain is top tier 😂
SO ONE MORE MODEL
Thorough Bench is a surprising thorough benchmark.
hollyyy no ads today
yup, white collar jobs are cooked
This is the first model where I cannot make out if it's real fireship or Fable speaking.
The 1,000-agent swarm is the part that stuck with me — a proof 7x shorter, while Opus 5.5 crashed at 26%. That's a very different kind of benchmark result.
Argon did nothing wrong
The Financial Times wrote that going from 2 to 1.5 is called an "increase". So math IS an opinion after all...???
We need to change the benchmark prompt every time, these models are learning.
the horse looks like ant eater and a horse had baby
The horse back limbs are like handles
I remember this channel for how technical it was and the joy that I felt when topics were explained. Now is just a AI news channel.
i wonder how gemini carbon will do in the horse benchmark
This is clearly AI voice
Are we in boiling pot with this model s ? As swe
"State of the art...in HORSE" 😂
I approve of all Secretariat worship.
The shock absorbers 😂😂
"Something Huang I can feel it" - Eminem
I find it hilarious that you are incentivising companies to optimise their models to create 3d horses
is this use tryna create rdr3 or what ??
i think we can all agree the horse benchmark should be become an industry standard
Jensen’s PR team is incredible. The entire conversation is about how he cheated on his wife with a hot young Chinese woman instead of about how he has been intentionally, illegally, and possibly treasonously exfiltrating top end GPUs and selling them to Chinese labs via his hot Chinese contact that he also sleeps with sometimes.
Imagine the Engineers at OpenAI and Anthropic watching this Horse Benchmark and they train their next model Specifically for this to success?
That's why USA told No to chinese Models.
Betteridge's Law
I'll vibecode mux
"claude cinematic universe" lol
with all this talk of horses I'm curious if fireship secretly plays umamusume
I want that horse disguise lol
Holy shit, its that minecraft 2 horse game notch created
I can't believe the technology nowadays, I have a realistic horse running on my computer 🐴
ThoroughBench and Matt Wolfe's BuseyBench are the only benchmarks that matter
horse benchmark is the only benchmark i care about now.
apparently Fable has never seen an actual horse head that looks like an anteater
Shock absorbers 😂
The model launch is the noise. The number to watch is the range in the amended filing: $70 to $78 a share for 825 million shares. That is why headlines built on the same range quote very different valuations: some use shares outstanding, others the fully diluted count. Before comparing it to any multiple, check which share count sits under the valuation.
Now this is a down to earth Vid from Fire guy... Fantastico...
Phar Lap is better than Secretariat don't @ me.
I'm waiting for this benchmark to be used in research
horse only goes 40 kph<br>it's an insult to racehorses thoroughbreds<br>5/10 max
I bet no AI can decipher Vayonich manuscript. For me that’s the BENCHMARK now.
Shock abosrber
I bet anthropic/openAi now have an horse department, to make a decent horse looking before delivering their models...
Nvidia CEO: Don't huang up on me
A swarm of 1,000 Fable agents finishing the math project in 30 hours with a proof 7 times shorter is wild. The catch is Opus 5.5 crashed out at 26% on the same setup, so that result says as much about the old model as the new one.
Your audio always make me think my speakers are blown out
Why the horse kinda ...
More hype videos for AI to keep the companies from plunging into the void 😂
I love the horse benchmark 😹😹😹😹
i thought it was 2 trillion ?
If you continue like that ai companies will start benchmarking their models on horsetinder
"anthropic suggests you should stick with opus 5 for most work and only swap to fable if you really need it", then why'd they give me 2 fable resets and only 1 weekly reset. it's a marketing gimmick i tell ya
That horse had me on a roll😭😭🤣🤣
If I were Anthropic, I'd have a hidden skill in claude code for Generating 3d models of horses.
LOL that chinese models quote taken completely out of context is so hilarious 😂😂😂
Y don't u change the name of this to ai report.
I think we need model capabilities and ecosystems which can verify the correctness of the built software over ever increasing code generation abilities. The ability to follow the guardrails and calling out when it’s unsure is needed more than anything else.
burning this domain for a horse is absolute cinema
So... no video about the honey pot? Comon! You can milk a video out of that. I believe you!
Huang horse all in one video.
Anthropic is a filthy company. <br>Cancel your subscriptions.
It only takes Haung mistake 😔
lol $122 per every person on earth
BRO THE HORSEBENCH HAD ME DYING 🤣🤣 ANOTHER HUGE GALLOP TOWARDS AGI
A horse with built-in shocks? very creative.
i only clicked to see the horse slop app
save? Anthropic is the best ai for common folk. Theyre going to skyrocket in a few years time
Why did anthropic IPO go to 1T from 2T? Educate please
Shock absorbers LMAOO
I got into electronics engineering because I was convinced AI would replace programming. This just feels like a bad joke now.
HorseBench 🐎
It's telling that, despite the clearly extreme compute used based on the footnotes of the press release, they made advances in just two extremely niche parts of science. People easily forget just how big the frontier of research is. The most impressive advance is in the speed up of the Fermat's last theorem formalization and Thoroughbench. But there's no way they can guarantee these are free of leakage. By Occam's razor...
Horse shock absorbers are pretty sick
I was expecting Fireship to have the most challenging AI benchmark in the industry... but here we are!
This is by far the funniest and most informative channel on the Internet. 5/5.
He's egoistic just like Kenpachi; trying to use logic to scale his power is just stupid, and I believe with his IQ, him being immersed in a new knowledge technique, he's probably going to power up in a way that will be overly stupid to comprehend
Always a shitpost that’s not a shitpost but an actual banger
He's egoistic just like Kenpachi; trying to use logic to scale his power is just stupid, and I believe with his IQ, him being immersed in a new knowledge technique, he's probably going to power up in a way that will be overly stupid to comprehend
I wonder if the model providers now run ThoroughBench prior to release to make sure they get good reviews from Fireship?
Chinese "models" huh, Mr. Huang?
Is horse tinder hosted yet?
🤬🤬🤬 Did you know that Trading 'funded' prop firm accounts are entirely simulated Or DEMO ? Even after qualifying, you trade with demo money rather than real capital. Furthermore, if you generate too much profit, firms often suspend your account under the pretext of rule violations , They just block your account. 🤬🤬
From this channel
