Super excited to announce seven new world-class MAI models today. They represent what we consider a new era in AI designed to keep you in control and on the frontier.
First is our text foundation model, MAI-Thinking-1, exceptionally strong on reasoning and SWE tasks.
- It’s a 35B active parameter MoE with a 256K context window. Independent human raters on Surge prefer it for overall quality in blind side-by-sides versus Sonnet 4.6, and it’s achieved 97% on AIME 2025, the key measure of its general-purpose reasoning abilities.
- It's at 53% on SWE Bench Pro, placing it right alongside Opus 4.6 on one of the toughest coding benchmarks.
- And since we co-designed our models with our own silicon, MAI-Thinking-1 is optimized on our MAIA 200 chip. Benchmarking head-to-head against the GB200, we see 30% better performance per dollar as well as a 1.4x performance-per-watt gain when running our MAI models on the MAIA 200 end-to-end.
Next is MAI-Image-2.5 and its Flash variant. Two super strong models now at #2 on the leaderboards, surpassing the score of Nano Banana 2 on image editing.
Last for now is MAI-Code-1-Flash, our new inference efficient coding model, especially tuned for VS Code and GitHub Copilot CLI.
- Code-1-Flash achieves 51% on SWE Bench Pro, despite having just 5B parameters, putting it closer to Haiku in size but cheaper in cost.
All of this is the foundation for Microsoft Frontier Tuning. It lets you customize our models to create custom, company-specific agents that only you control. You can make our model, your model. Your data. Your agents. Your moat.
Early adopters are already seeing a difference. When we tuned our models for McKinsey’s tasks, MAI delivered the highest win rate, outperforming GPT-5.5 on quality, while being 10x lower on cost.
Also really excited to be collaborating with the amazing team at Mayo Clinic to jointly train a new frontier AI model for healthcare.
Our announcements today mark another milestone on the road to humanist superintelligence. You can learn more and about our other new models in our latest blog: https://t.co/v65eop5Ixq
Twenty years ago, @Microsoft broke ground on our first data center in Quincy, Washington. At a time when so many small towns have been shrinking, Quincy has grown, and the partnership behind that story is one worth telling. Watch how a farm town in Central Washington helped build the cloud, and built a thriving future of its own.
Yes, the United States has the most progressive tax system in the world. The top 1% pay 40% of taxes, the bottom 50% pay 3% of taxes. We can make it even more progressive by zeroing out taxes on the bottom half. It’s a small amount of the total tax revenue but very meaningful to people in this group.
WorkIQ is an incredible benefit of being in the MSFT ecosystem.
You don’t realize it until you ask about other documents or emails and the agent just knows.
We see our home planet as a whole, lit up in spectacular blues and browns. A green aurora even lights up the atmosphere. That's us, together, watching as our astronauts make their journey to the Moon.
If you're building with GitHub Agentic Workflows, security is baked into the foundation. 🔐
The architecture is set up around three core principles:
🛡️ Isolation
🛡️ Constrained outputs
🛡️ Comprehensive logging
Here's how we engineered it to be secure by design from day one. ⬇️
https://t.co/tmJDgqlIJi
Custom orders of the Tesla Model S & X have come to an end. All that’s left are some in inventory.
We will have an official ceremony to mark the ending of an era. I love those cars.
This was me at production launch 14 years ago:
If the AI revolution doesn’t result in 10x higher quality software when being driven by experienced engineers, I’m gonna be so disappointed in us. We should be “over cooking” the things we undercooked before because of the cost.
The real story though, is that we propose a program where companies must invest deeply in our EiC pipeline, not just hiring juniors but giving them formal *preceptors* (modeled after nursing) such that we create a strong pipeline of new senior engineers to take on the coming industry challenges.
I’ve been on a CGM since they were invented, and I created the first portable glucose system in the 90s on a palm pilot.
Here is a part you’re not gonna wanna hear - if you’re not a diabetic, you don’t need a continuous glucose meter.
You’re not gonna learn anything. There is no glucose spike when you eat this or that - you’re normally blood sugared person. A normal person’s blood sugar can go as high as 150 or more for minutes at a time after eating pizza or sugared soda. This is not unique.
They will happily sell you a meter, and they will happily take hundreds of your dollars, and they will also stretch the Y axis to make it look like your blood sugar “spikes to 120 when you eat grapes.” then you’ll say I learned so much about my body and what happens when I eat this or that or the other thing.
But that’s what a normal pancreas does.
I haven’t had a working pancreas for almost 35 years.
Enjoy the pancreas you have and there’s absolutely nothing that you’re gonna learn from a continuous glucose meter unless you are either a type one or a type two diabetic, or you’re a prediabetic with a genetic disposition and your doctor recommends it.
I apologize for the rant, but they are selling continuous blood sugar meters to humans that don’t need them when there are actual diabetics that do.
All that said that Dexcom Stello will do what you want
People should not be judging AI-augmented coding by “1 shots.”
If someone told you that their model did a “one shot of Minecraft” and they’re impressed by that, you need to consider how much semantic heavy lifting the word “Minecraft” is doing in that prompt.
Ask them to one shot Minecraft without using the word Minecraft.
It’s not trivial to one shot something unique, because programming is the art of making the ambiguous incredibly specific through sculpting. AI sculpting is less about vibes and more about finding the specificity you want and keeping the system stable through changes. Good SDLC practices still matter, historical context still matters, and knowing how things work matters, shout out to @Grady_Booch.
It’s a cool party trick to one shot Mario Brothers or space invaders, but then you’ll end up with the most mid version of both. Literally mid. You’ll get the statistical fat part of the Bell curve version of these mythical games. You’re telling the model to close its eyes and draw the face of these games from memory.
As high-level programming cedes way to the prose compiler, making your goals and specs well understood to the ambiguity loop and showing good judgment is going to matter more than ever. Consider all of your words and make sure that certain words aren’t carrying all the semantic load, hidden or otherwise.
Our newest AI accelerator Maia 200 is now online in Azure.
Designed for industry-leading inference efficiency, it delivers 30% better performance per dollar than current systems.
And with 10+ PFLOPS FP4 throughput, ~5 PFLOPS FP8, and 216GB HBM3e with 7TB/s of memory bandwidth it's optimized for large-scale AI workloads.
It joins our broader portfolio of CPUs, GPUs, and custom accelerators, giving customers more options to run advanced AI workloads faster and more cost-effectively on Azure.
@BambulabGlobal I was never a fan of painting the models, that’s why I got the X1C. But oh man the purge waste! With H2C, I can finally print Christmas trinkets in full color without filament waste. Especially looking for printing a full color Santa!
With H2C, the limit is no longer the colors you can paint by hand, but the ideas you can imagine.🎨
Up to 7 colors in a single print and 24 in total, meaning those hours of hand-painting might finally become history. 🖌️ To celebrate the launch, we’re giving away our first H2C AMS Combo! 🎁
Tell us how H2C would change the way you bring your multi-color ideas to life:
1️⃣ Follow us
2️⃣ Like & Repost 🔁
3️⃣ Comment: Which project will you finally be able to print instead of hand-paint? 💬
📅 One winner will be randomly selected and announced on December 2.
H2C turns “I wish I could print that” into reality. Let’s paint the future, one print at a time. ✨
https://t.co/1Jr1iGBJTx
#H2C #BambuLab #3DPrinting #PrintBeyondPaint