Top
Best
New

Posted by EvanZhouDev 1 day ago

MAI-Code-1-Flash(microsoft.ai)
https://microsoft.ai/models/mai-code-1-flash/

https://microsoft.ai/pdf/MAI-Code-1-Flash-Model-Card.PDF

Launching seven new MAI models: https://microsoft.ai/news/building-a-hillclimbing-machine-la...

520 points | 245 commentspage 4
giancarlostoro 23 hours ago|
Mark Zuckerberg must be in crisis. Microsoft releasing models that compete with Claude's models. Meanwhile the only thing anyone knows about Mark's models is that they help you get hacked more easily.
ggcr 23 hours ago||
Meta recently launched Muse Spark [1] and they themselves compare against Claude Opus 4.6 Max.

Here Microsoft is comparing against Claude Haiku, the smallest and least capable model from Anthropic.

[1] https://ai.meta.com/blog/introducing-muse-spark-msl/

hashmap 21 hours ago||
i have had good results adding muse spark's contemplate mode as a roundtabler for complex questions. but you cant turn off their data ingestion for training so that is a shame.
yuppiepuppie 23 hours ago||
Wait… I think he has moltbook IP as well that he can scale up.

Seriously tho, wtf is going on over at Meta? Anyone working there currently want to describe the vibe of the org when it comes to being a frontier company?

giancarlostoro 23 hours ago||
I don't understand his plan, if I were him I'd either have just gone all in on making RAM which would become very lucrative, or would have focused on building programming models. They've built some key open source technologies, but its as if Mark Zuckerberg cannot run anything that isn't a social media company / project.
mmaunder 1 day ago||
You lost me at forced scrolling. Ugh!
Tepix 1 day ago|
From https://news.ycombinator.com/newsguidelines.html

Please don't complain about tangential annoyances—e.g. article or website formats, name collisions, or back-button breakage. They're too common to be interesting.

bguberfain 1 day ago||
It is good to se big companies like Microsoft launching LLMs. They have large amount of compute power and good scientists to create useful models.
ComputerGuru 1 day ago|
Microsoft has been releasing LLMs for years.
ipsum2 1 day ago|||
Sort of. Phi models were just trained on GPT outputs though.
kingstnap 23 hours ago|||
For those that don't know about this. Phi was announced with a paper called "Textbooks are all you need". What they did was use GPT 3.5 and created synthetic textbook chapters and exercises.

They also did some more interesting work like showing very small models can be coherent as long as you have very simple children's book style training data (TinyStories is pretty famous).

Lots of these ideas are still used. Learning facts at scale with active reading is an ICLR 2026 paper from Meta AI that does a lot of similar work.

not_a_bot_4sho 23 hours ago|||
By design. The whole point of Phi is the "textbooks is all you need" theory on curated training data, as opposed to kitchen sinks.
jwitthuhn 1 day ago||||
And occasionally un-releasing them like with WizardLM.
lemonish97 1 day ago|||
They were mostly distilled or fine-tuned OAI models.
Havoc 23 hours ago||
huh? The granite series isn't distilled
wirybeige 23 hours ago||
Granite is IBM
Havoc 11 hours ago||
Ah snap. You’re right of course
GaryBluto 23 hours ago||
What's with the lack of Microsoft design language on the website? It's painfully obvious they're trying to emulate Anthropic's style here and it looks tacky.
foltik 23 hours ago||
Definitely vibed microslop, the giveaway is the broken header and scrolling on mobile.
lanyard-textile 22 hours ago||
The broken header is an incredible distraction. I can't believe this slipped through.
shrinks99 21 hours ago|||
Brand guidelines and web design pretty much don't exist any more as far as I can tell. Gotta get it out yesteday, and the only way to do that is vibe coding, styling be damned.
Handy-Man 22 hours ago|||
That's neither Microsoft nor Anthropic design. It's from their acquisition of Inflection AI. Even Copilot mobile app design is basically what was Inflection's design
singhkays 22 hours ago||
I've always wondered where Consumer CoPilot's design language was from.

If you watch the Build keynote with Satya, you'll notice that the design of the slides changed to Serif typography and warmer colors when Mustafa/Microsoft AI segment came on which was completely different from the rest of the keynote. Now it makes sense!

not_a_bot_4sho 20 hours ago||
Tangentially, I've seen a couple people use "CoPilot" here instead of "Copilot" (the actual product name).

Where does the Pascal case inspired variant come from? Is it a reference to something? Is it like "M$" was used back in the days?

flumpcakes 21 hours ago|||
Thank you! This website is dreadful for accessibility and usability.
i_have_an_idea 23 hours ago|||
maybe it was coded by Claude
winfredJa 23 hours ago|||
i think it is AI generated.
petercooper 22 hours ago||
"It’s not just smarter; it’s leaner"
gedy 22 hours ago|||
This is needlessly embarrassing, seems like a small thing, but it makes them look... desperate?
stringfood 23 hours ago||
A little to minimalist - only a few hundred words on entire page!
hootz 1 day ago||
I'd love to see a tokens per second metric. I always prioritize speed over raw intelligence for flash models.
throwaw12 1 day ago|
> I always prioritize speed over raw intelligence for flash models.

This model might have a perfect speed:

    for i in range(100):
      print(random.choices(words))
OsrsNeedsf2P 23 hours ago||
Leave it long enough, and it'll print the work of Shakespear!
ruined 22 hours ago||
wtf are they doing to the scroll on that page
elpakal 20 hours ago|
even worse on mobile
striking 1 day ago||
To be clear about the size of the model: MAI-Code-1-Flash is 137B A5B.
halapro 15 hours ago||
In a few languages MAI means no/never, so it's an apt name for a Microsoft offering.
jeffrallen 14 hours ago|
MAI? Ma die, mai!

(gestures wildly while changing lanes in his Fiat 500)

notenkidev 14 hours ago|
Curious how this handles token cost visibility. One of the biggest pain points with AI coding tools right now is having no idea what you're actually spending per project.
More comments...