This article is a good step toward understanding how things really work. However, quite a few factors related to the experimental setup need to be taken into account if you want to draw more reliable conclusions:
1. The optimal shape of each architecture may differ under compute-optimal settings.
2. The optimal initialization of each parameter may differ across architectures.
3. Different architectures may respond quite differently to different optimizers.
4. Differences in training stability and the health of “internal metrics” can lead to drastically different results, even for the same architecture.
5. Architectures may respond differently to varying degrees of overtraining (tpp).
6. Different architectures may respond quite differently to different data distributions (i.e., different learning objectives).
7. The difficulty of scaling up can vary considerably across architectures.
There are many more nuances that need to be carefully addressed to obtain consistent results.
Show more
Finally releasing the results of the arch ablations I was running with bux over the weekend
Keep pushing the frontier of open weight models
Introducing Kimi K3.1 Pro: our most capable 3T-class open-weight model.
🔹 1M-token context, 2.8T total / 104B active params (MoE), Multimodal.
🔹 Full-project coding: games in Unreal, Unity & Godot.
🔹 Excels at parametric 3D modeling in Blender & CAD.
🔹 More reliable at finance work: Accurate modeling with hallucination down by 65%.
🔹 Matching the strongest proprietary models across coding, agentic and knowledge benchmarks.
Kimi K3.1 Pro and K3.1 Pro Highspeed (2–3× faster) are now live on Kimi Work, Kimi Code, and the Kimi API.
Open weights by Nov 4, 2026.
🔗 API:
🔗 Tech blog:
Show more
Introducing Embodied First Exam:
One line of code to run coding agents in 1,090 embodied tasks across 31 simulation suites and 44 robots.
Can your agent go far beyond pick-and-place:
Puzzles with spatial constraints;
Olympic sports with multi-robot collab;
Real-time eval, where the world won't pause;
Dynamic, agile, whole-body control;
AutoResearch with sim2sim
Try your own agent:
harbor run -d embodied-first-exam -a -m
Show more
i asked GPT 6 Pro and Fable 5.1 to rank all discoveries in the last three years
🔵 for Human discovered
🔴 for AI discovered (before October 6th)
🟢 for AI discovered from OpenAI/math repo
81% of them have been released today. wtf
Show more
Quasi-RH?!?!???! Are you kidding me? If a human did this, it would be an instant Fields Medal, no questions asked.
RH says zeta has no zeros in Re(s)>1/2. The best we had until a second ago was a region that got thinner and thinner the higher up the imaginary axis you go. I thought maybe they’d fatten that up a bit, that’d be a massive breakthrough. But no. They got a zero free strip!!!! Insane
Show more
what a time to be alive...
We’re sharing a solution to the Navier-Stokes Millennium Prize Problem, one of the deepest problems at the frontier of mathematics.
The proof was produced by a group of agents, using an OpenAI next-generation model significantly more capable than GPT-6 Astra.
The problem concerns whether the description of smooth three-dimensional fluid motion modeled by the Navier-Stokes equations can break down. It has remained unresolved for roughly 90 years.
Show more
I havent updated this benchmark in a while. Astra completely saturates my spatial reasoning eval. I am at a loss for words, and i'm declaring LLM vision solved. Every so often when a new model came out I would test it on a sample question and it failed. I tested Astra and it kept getting answers correct, so I evaluated it...
Show more
Deconstructing Scaling Laws: The Triad of Optimization, Architecture, and Data
This is so creative, I love this video!!
Cool
Done.
Will do
true. it is the same model
@kimmonismus moonshot confirmed during a Reddit AMA today that the released MXFP4 quantization is the exact same format used for their hosted API, so self-hosted inference quality should match what people have already been testing since the July 16 launch.
Show more
So many technical details. Proud of what the team shipped.
Releasing the model weights and technical report of Kimi K3.
Kimi K3 is our most capable model: a 2.8T MoE model with native visual understanding and a 1M-token context window.
New model architecture: 2.5x the intelligence per unit of compute, not just more params.
Alongside Kimi K3, we're opening up more of the stack behind it — high-performance attention kernels, MoE communication library, and infrastructure for running agent environments at scale.
Model weights:
Tech report:
Tech blog:
Show more
For my first post, I’m sharing a letter @NVIDIA signed on why open models matter.
AI will transform every industry, power every company, and be built by every country.
Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty.
The world needs both frontier closed models and frontier open models.
Show more
Good. Could be even better.
BREAKING: Kimi K3 by @Kimi_Moonshot is 1st overall on 3D Design with an Elo of 1450.
This is a 6 position and 108 Elo jump from @Kimi_Moonshot's previous model, Kimi K2.6. This performance puts Kimi K2.6 82 Elo ahead of Claude Fable 5 by @AnthropicAI in 2nd and 87 Elo ahead of GLM 5.2 by @Zai_org in 3rd.
Congratulations to the @Kimi_Moonshot team on this accomplishment!
Show more
true
Prediction is compression. Same thing.
BREAKING: Kimi K3 by @Kimi_Moonshot is officially 1st on Frontend Web App Arena by DesignArena
With an Elo of 1326, this open-weight model leads the way, ahead of Fable 5, Sonnet 5, and Opus 4.8 by @AnthropicAI
Huge congrats to the @Kimi_Moonshot team for this achievement!
Show more
we're working extremely hard with our partners to open the weights as soon as possible
So cool!
你敢信？ 这是我单靠一张嘴撸出来的啊！
你可要知道这个东西之前JigSpace就靠这个发家致富的啊～
太特么细腻和细节到位了…
I’m a huge Hollow Knight fan working at Kimi. Before K3 was released, I used Hollow Knight assets to build a game where the Knight battles Kimi.
Try it here:
You’re also very welcome to check out the other projects on K399. They’re all a lot of fun!
Show more
