Rendered at 22:29:35 GMT+0000 (Coordinated Universal Time) with Cloudflare Workers.
keeda 3 hours ago [-]
> Second, clean data. MAI-Thinking-1 was trained on clean and appropriately licensed data, with AI-generated content excluded from pre-training. This matters for quality, provenance, and control. If we cannot account for what shaped a model, we cannot fully understand its behavior or credibly improve it.
Shots fired?
It would be interesting to see how far "clean data" can go on the scaling laws.
foresterre 1 hours ago [-]
I would really like to see what "appropriately licensed data" means. Cannot imagine they didn't copy all open repo's on GitHub, and can't imagine they asked for permission, or are reproducing license texts from these repo's now. It sounds hand wavy.
P.S. A fairly basic website otherwise, but it unfortunately seems to be hacking scroll for no good reason.
stingraycharles 1 hours ago [-]
I assume they took the actual repos’ licenses info account. I don’t understand why they should ask for permission when the license would already allow for it.
foresterre 22 minutes ago [-]
Almost all licenses have requirements to redistribute copies of the work, or derivatives thereof. Even permissive licenses do. It's very little to ask when open source dev's provided thousands of hours of free work.
For example, the Apache 2.0 license requires in just 4.c:
You must retain, in the Source form of any Derivative Works that You distribute, all copyright, patent, trademark, and attribution notices from the Source form of the Work, excluding those notices that do not pertain to any part of the Derivative Works;
Just because they're tokenized and transformed into a probabilistic mapping, doesn't suddenly mean that they weren't copied.
I find it morally unethical that they (likely) just ingest IP of all open source repo's without asking, but also importantly without any attribution.
Let me also note that I'm not against LLM's in general. But I do think training on open source must be opt-in, and I look forward to a world with actually ethical, and traceable (i.e. on what they were trained on, like a bill of materials (BOM)), models.
rocqua 59 minutes ago [-]
Which licenses allow usage for training? MIT, BSD, etc likely do. But I would expect it gets weird for all the various copyleft licences.
cortesoft 33 minutes ago [-]
Why would it get weird for those?
rzmmm 27 minutes ago [-]
Theoretically it mandates that derivative works use same license but it's unclear if that applies to LLM outputs.
supermdguy 56 minutes ago [-]
It's interesting because their last model series (Phi) was based around the thesis that high-quality synthetic data is better than a large pre-training corpus.
vdfs 2 hours ago [-]
I doubt any lab would say otherwise, they all _claim_ to use licensed data
keeda 2 hours ago [-]
Maybe, but Microsoft, through their partnership with OpenAI, is already involved in major copyright lawsuits. That is probably a driving force for this move, actually... I doubt they would want to tempt fate while those lawsuits are on-going.
2 hours ago [-]
andai 27 minutes ago [-]
Interesting. Wasn't their previous attempt (Phi) trained mostly on synthetic data?
swalsh 44 minutes ago [-]
I'd assume it's not up to par with Qwen-3.5 then, which has been distilling Claude, and the quality of the model is probably a direct result of that.
vanuatu 24 minutes ago [-]
all the labs "clean" their pretraining data, and you can have your pretraining data to be minimally ai generated but also spam synthetic post-training data
onlyrealcuzzo 3 hours ago [-]
I'm interested how much "Clean Data" is synthetic data from "unclean" models...
bicx 2 hours ago [-]
So, laundered data?
ertgbnm 2 hours ago [-]
> with AI-generated content excluded from pre-training.
> without distillation from third-party models
sounds like zero unless they are lying.
zamalek 2 hours ago [-]
> with AI-generated content excluded from pre-training.
Though this is largely impossible these days, unless they pre-trained on pre-AI era data.
stymaar 13 minutes ago [-]
That could be. Just use pre-training for language understanding and let the post-training on synthetic data do the heavy lifting.
2 hours ago [-]
saghm 1 hours ago [-]
"how many of those shapes are rectangles?" "sounds like zero unless they are squares"
Adding "unless" to a statement makes it vacuous if the latter clause is weaker than the first clause. I find it hard to believe that a company willing to violate licenses would have scruples about lying about it.
rocqua 55 minutes ago [-]
Not vacuous, but tautological.
Which is different, because tautologies can actually be quite directly informative. Whereas vacuous truths tend to be oblique.
Also, “Microsoft is lying” is not a logically stronger statement, because they might be lying about something other than whether they distilled or trained on AI output.
chongli 1 hours ago [-]
Adding "unless" to a statement makes it vacuous if the latter clause is weaker than the first clause
I think that's the point. "How do I say they're lying without outright saying they're lying?"
It's a common rhetorical trick.
xavriley 2 hours ago [-]
“ We trained it from the ground up on enterprise grade, clean and commercially licensed data, without distillation from third-party models.”
It's good there is a new player on the market, I take benchmark tables with a grain of salt, however. Speaking about model presentation it's funny to see how clearly their website is inspired by other AI company blogs with extra innovation of hijacked scrollbar.
1 hours ago [-]
Alifatisk 1 hours ago [-]
> MAI-Thinking-1 is built with enterprise readiness in mind. It supports long context with a 256k token window
Isn’t 1M becoming the norm?
vb-8448 57 minutes ago [-]
1M it's only marketing, in my experience above 150k quality noticeable drops.
Claude code will suggest you to start a new session or compact if you go above 100k.
stingraycharles 1 hours ago [-]
Yes it is, but I can imagine that they want to start out a bit smaller to see how well things scale, and/or did not yet have the time to work on optimizing for the large context windows.
droidjj 1 hours ago [-]
I struggle to get quality results from the frontier models at contexts > 256k anyway.
stingraycharles 31 minutes ago [-]
Yup, same experience, it’s because the attention basically has exponential complexity. So at large context windows, they need to compress the attention (eg group multiple tokens together), when then leads to loss in accuracy.
It’s almost always better to keep your context windows small.
jampekka 50 minutes ago [-]
The benchmarks are a bit of a disaster? It's at about DeepSeek V3.2 level, but with about 50% more parameters. Loses handily to the also smaller GLM-5.1, and even worse to the similarly sized Kimi K2.6.
usef- 35 minutes ago [-]
They claim to not be training to the benchmarks at all. It'll be interesting to see how it stacks up in actual use.
sailingparrot 39 minutes ago [-]
Yes and no.
Yes from a user PoV, I don't really see a great reason to use this other than for enterprises that care about using a model not trained on copyrighted data (not sure what the market really is for this anymore, feels like this concern has been forgotten by most customers.)
From a strategic PoV for MS, all the model you cited are distilling GPT/Claude/Gemini and wouldn't be anywhere as good as they are without this distillation, which in turn means you are dependent on OAI/Anthropic/G first shipping a good model to generate data for your training. This MAI model is trained from scratch with no synthetic data or distillation. So in term of benchmark its obviously much harder to get strong score and thus not a disaster if they can keep on improving.
pixeldash928 4 hours ago [-]
Looks like the OAI divergence is finally taking place. Seems like the comparisons are mainly with Opus 4.6 and GPT 5.4 though. Still, exciting to see a new frontier player.
i_have_an_idea 2 hours ago [-]
Is it a frontier player though, or perhaps a new benchmaxxed model? People were saying similar things about Grok but it ultimately amounted to little.
wasabi991011 2 hours ago [-]
"preferred by humans over Sonnet 4.6" makes it pretty clearly not benchmaxxed though.
At least when you define benchmaxxed as "good in benchmarks but not human preference".
dude250711 1 hours ago [-]
Post 4.6 Anthropic models do not exactly have a stellar reputation, so that choice is smart.
> MAI-Thinking-1 is a 35B-active, ~1T-total parameters, sparse Mixture of Experts model, a smaller inference footprint than much larger models.
This seemingly nonsensical sentence (of course this will have a smaller inference footprint than larger models) suggests this model's competitors have larger inference footprints and total parameter sizes.
BeetleB 2 hours ago [-]
Based on the first table, why would I pick this over GLM?
missedthecue 1 hours ago [-]
Because your employer might make you exclusively use enterprise copilot.
BeetleB 59 minutes ago [-]
As long as my employer is footing the bill, fine.
For personal stuff this release is not noteworthy.
hartator 2 hours ago [-]
I like it so much when a website hijacks the way my scroll works. This is truly innovative.
campital 23 minutes ago [-]
Yeah, you might get disoriented and throw up if they didn't smooth it out.
lordmauve 2 hours ago [-]
We need to see DeepSWE scores. SWE Bench Pro is junk.
kaicianflone 1 hours ago [-]
Is that a pretext zoom effect when changing screen dimensions? Very cool.
wmf 2 hours ago [-]
At least there shouldn't be any complaints about benchmaxing this time.
i_have_an_idea 2 hours ago [-]
Just because it is performing rather poorly by comparison, it doesn’t mean it isn’t benchmaxxed. It can still be worse than it appears.
wasabi991011 2 hours ago [-]
It isn't benchmaxxed because they are using human preference as an evaluation.
gigatexal 58 minutes ago [-]
Anyone believing those benchmark numbers from a 35B model?
jeffdn 55 minutes ago [-]
It says right at the top, 35B active, 1T total.
kstenerud 2 hours ago [-]
They've hijacked scrolling. They've hijacked the spacebar. It flickers like crazy when I try to move through the article. Trying to get through it is an exercise in madness.
t-sauer 2 hours ago [-]
I do not understand how scroll hijacking is still a thing. Who thinks this is a better experience?
maelito 2 hours ago [-]
Designers.
bensyverson 10 minutes ago [-]
As a designer, let me tell you: scroll jacking is not good design
AirMax98 2 hours ago [-]
I normally don't comment on matters of taste like this, but wow this is brutal. It's like someone threw the site in a vat of molasses.
grassfedgeek 1 hours ago [-]
Even without flicker it is very distracting. Why do people think this is a good idea?
aniceperson 2 hours ago [-]
there is also a gap between the header and the top of the page... they should ask the ai to make it better a few more times...
blisstonia 2 hours ago [-]
I gave up after the first scroll.
vcryan 2 hours ago [-]
It really looks like they used Claude to design this webpage. I guess the color taupe it the marker of good AI today.
Handy-Man 1 hours ago [-]
Inflection AI
bossyTeacher 2 hours ago [-]
7 modes launched. 5 models in the dropdown. Only 4 actually usable :(
About time Microsoft joined the fray. After the OpenAI divorce, it really looked like Microsoft was going to become another Uber.
giancarlostoro 2 hours ago [-]
They still own 27% of OpenAI, this IPO will feed them a lot of easy cash.
2 hours ago [-]
simjnd 4 hours ago [-]
Absolutely disgusting scroll jacking, even when "Accessibility mode" is turned on
dang 3 hours ago [-]
I'm sure most of us agree, but:
"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."
Shots fired?
It would be interesting to see how far "clean data" can go on the scaling laws.
P.S. A fairly basic website otherwise, but it unfortunately seems to be hacking scroll for no good reason.
For example, the Apache 2.0 license requires in just 4.c:
Just because they're tokenized and transformed into a probabilistic mapping, doesn't suddenly mean that they weren't copied.I find it morally unethical that they (likely) just ingest IP of all open source repo's without asking, but also importantly without any attribution.
Let me also note that I'm not against LLM's in general. But I do think training on open source must be opt-in, and I look forward to a world with actually ethical, and traceable (i.e. on what they were trained on, like a bill of materials (BOM)), models.
> without distillation from third-party models
sounds like zero unless they are lying.
Though this is largely impossible these days, unless they pre-trained on pre-AI era data.
Adding "unless" to a statement makes it vacuous if the latter clause is weaker than the first clause. I find it hard to believe that a company willing to violate licenses would have scruples about lying about it.
Also, “Microsoft is lying” is not a logically stronger statement, because they might be lying about something other than whether they distilled or trained on AI output.
I think that's the point. "How do I say they're lying without outright saying they're lying?"
It's a common rhetorical trick.
Isn’t 1M becoming the norm?
Claude code will suggest you to start a new session or compact if you go above 100k.
It’s almost always better to keep your context windows small.
From a strategic PoV for MS, all the model you cited are distilling GPT/Claude/Gemini and wouldn't be anywhere as good as they are without this distillation, which in turn means you are dependent on OAI/Anthropic/G first shipping a good model to generate data for your training. This MAI model is trained from scratch with no synthetic data or distillation. So in term of benchmark its obviously much harder to get strong score and thus not a disaster if they can keep on improving.
At least when you define benchmaxxed as "good in benchmarks but not human preference".
MAI-Code-1-Flash - https://news.ycombinator.com/item?id=48374466 - June 2026 (131 comments)
This seemingly nonsensical sentence (of course this will have a smaller inference footprint than larger models) suggests this model's competitors have larger inference footprints and total parameter sizes.
For personal stuff this release is not noteworthy.
About time Microsoft joined the fray. After the OpenAI divorce, it really looked like Microsoft was going to become another Uber.
"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."
https://news.ycombinator.com/newsguidelines.html