bycloud

I make youtube videos on cool AI research /// AI papers newsletter /// paper recap @TheAITimeline ///

[bycloud @bycloudai]

@elliotarledge @calebfoundry my goat

[bycloud @bycloudai]

pov you’re too deep into papers and when u see this you know what it’s exactly referring to

[bycloud @bycloudai]

Microsoft's new MAI-2, a new pre-trained base model they made.

What I found interesting:
- used muon and the loss is soo smooth
- didnt care about expert imbalance?!
- same pre-training quality as DeepSeek-V4-Pro-Base with 4.2-12.3x fewer FLOPs (omg)
- The entire 2.8T-parameter model was pre-trained in 38 days across 16k GB200s
- 21% model FLOPs utilization
- 89% goodput
- 50 trillion training tokens

[bycloud @bycloudai]

https://t.co/8sS2TR9XTT

[bycloud @bycloudai]

hey at least they shared a paper

[bycloud @bycloudai]

so you’re telling me you distilled a monster and it’s still a monster?

[Claude @claudeai]

Introducing Claude Haiku 5.5: the cheapest, fastest, and most capable small model we’ve ever released.

On average, it costs around 75% less to run than Claude Haiku 4.5.

[bycloud @bycloudai]

@zhomag thought it was around 30USD so i dont need to pay 100USD

turns out i paid 1k instead

[bycloud @bycloudai]

i thought this was in my own currency (so $1000 is ~30USD) but it's in USD?!?

my god, this $1000 option is diabolical and they wouldnt let me refund it :(

I know I shouldve double checked but my session was running out so I rushed it... Any help would be much appreciated

[bycloud @bycloudai]

@Antho8793 @marginference unfortunate...

[bycloud @bycloudai]

@marginference unfortunately it did not

I tried within 10 mins and got rejected 😅

I turned off the credits usage too and left it there lol...

[bycloud @bycloudai]

Instead of chasing viral news headline, Xiaomi literally made a dedicated hack agent to probe the environments as hard as possible before letting LLMs to train in the environments.

"continued this process until the hack agent could no longer find a successful exploit in any of the environments"

If Chinese labs can do it, why can't the other frontier labs do it? Makes you wonder if they are just being reckless or its simply some viral marketing stunt

This is open model SoTA btw, finding exploits is def possible, it's just if you want to "let" it escape or not

[bycloud @bycloudai]

so much to learn from this dashboard

[Fuli Luo @_LuoFuli]

Nearly half a year of silence. We spent it studying one problem: how far RL can scale.

MiMo-V2.6 is in the middle of its RL run right now. Three things we scaled: compute (~2B tokens per step, 1568 prompts × 16 rollouts, fully async), environments and harnesses (multi-task agentic RL, mixed across multiple harnesses in one run), and grader compute (agentic in-group credit assignment, with test-case and rubric-based rewards). We'll open-source the details piece by piece over the coming weeks.

Streaming the run: https://t.co/ZSxahzJRju

[bycloud @bycloudai]

@secemp9 lowk was thinking diffusion too

[bycloud @bycloudai]

this is tinkling the part of my brain where i wanna find out what architecture this is

[Diogo Almeida @CompleteSkeptic]

After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI?

I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev

• 20-200x faster
• 40-400x cheaper (w/ output tokens free)
• Frontier composable intelligence optimized for decisions

AFAICT the shortest path to AI-based economic revolution

[bycloud @bycloudai]

@mreflow @smoothmedia_ @davidheuff congrats!!!

[bycloud @bycloudai]

@calebfoundry LETS GO CALEB🎉🎉

[bycloud @bycloudai]

we are at a stage where 99.99% of the people have no idea what breakthrough LLM is making now

[Noam Brown @polynoamial]

An internal version of Astra, @OpenAI’s next major model family, solved 10 major open problems in mathematics, quantum complexity, and theoretical computer science.

We believe it will be a major step for scientific reasoning. https://t.co/iP6cyheZ7i

[bycloud @bycloudai]

@korigero blue stands for muse

[bycloud @bycloudai]

i guess 53.4% > 53.5% now

[Claude @claudeai]

On several coding and knowledge work evaluations, Opus 5 is the new state-of-the-art:

[bycloud @bycloudai]

Why do I feel like they’re going to ditch the “Fable/Mythos” brand names now because those names are too cursed?

And WHAT IF they used the Fable/Mythos branding on purpose as an experiment to see how people & governments would react, and if worst came to worst, they could just throw that branding under the bus?

[Claude @claudeai]

On several coding and knowledge work evaluations, Opus 5 is the new state-of-the-art: