NEW KIMI AI | Coding with a Multimodal Long-Context Brain

My Account

Mijn Account

Mon Compte

Mein Konto

我的帐户

Inloggen

Connexion

Anmeldung

Acceso

NEW KIMI AI | Coding with a Multimodal Long-Context Brain

Kimi K2 is the large language model series developed by Moonshot AI team

Kimi K2 is a state-of-the-art mixture-of-experts (MoE) language model with 32 billion activated parameters and 1 trillion total parameters. Trained with the Muon optimizer, Kimi K2 achieves exceptional performance across frontier knowledge, reasoning, and coding tasks while being meticulously optimized for agentic capabilities.

Key Features

Large-Scale Training: Pre-trained a 1T parameter MoE model on 15.5T tokens with zero training instability.

MuonClip Optimizer: We apply the Muon optimizer to an unprecedented scale, and develop novel optimization techniques to resolve instabilities while scaling up.

Agentic Intelligence: Specifically designed for tool use, reasoning, and autonomous problem-solving.

Model Variants

Kimi-K2-Base: The foundation model, a strong start for researchers and builders who want full control for fine-tuning and custom solutions.

Kimi-K2-Instruct: The post-trained model best for drop-in, general-purpose chat and agentic experiences. It is a reflex-grade model without long thinking.

https://www.kimi.com

https://aimode.co/app/kimi

Architecture	Mixture-of-Experts (MoE)
Total Parameters	1T
Activated Parameters	32B
Number of Layers (Dense layer included)	61
Attention Hidden Dimension	7168
MoE Hidden Dimension (per Expert)	2048
Number of Attention Heads	64
Number of Experts	384
Selected Experts per Token	8
Number of Shared Experts	1
Vocabulary Size0	160K
Context Length	128K
Attention Mechanism	MLA
Activation Function	SwiGLU

Wes Roth

wes-roth

NEW KIMI AI | Coding with a Multimodal Long-Context Brain

AI News | X-Omni | ChatGPT | Hera | GLM-4.5 | Qwen3 | Gemini

Matt Williams | New Ollama UI, The Game Changer for local AI

TechLinked | ChatGPT Was Leaking The Whole Time…

AI News | ChatGPT Agent | Grok 4 | Opal | Unitree | Chimera

OpenAI | Major Update to ChatGPT, Study Mode and AI Agents

Superintelligence Is Now In Sight | Mark Zuckerberg's Master Plan

Wan2.2 | The Best Free and Open Source AI Video Generator

Zhipu GLM-4.5 | The Smartest Open Source Model Ever Built?

ASI-Arch | This AI Autonomously Designs New AI Models

Ai Search | DAViD | SeC | Yume | ChatLLM | Qwen3 | Diffuman4D

AI News Videos Blur Line Between Real and Fake Reports

Matt Wolfe | GPT-5 Launch Date, Face Stealing Apps and Baby Grok

I Convinced Grok God Is Real Using Math, Science, and Logic

Creator Magic | Google Just Released Opal, The AI App Builder

AI Researchers Shocked as Models Quietly Learn to be Evil

HiDream-E1 | The #1 Free and Open Source AI Image Editor

QWEN 3 CODER is Unleashed | Its Better Than KIMI K2

AI Explained | How Not to Read a Headline on AI

The World’s First Interview With Elon Musk’s Grok 4 Heavy

Pusa | Spatial Tracker | ChatLLM | Kimi K2 | Epona | GPT Agents

Wes Roth | ChatGPT Agent is Out of Control

MattVidPro | Full Body AI Acting and Open Source Slam Dunks

Author David Baldacci Rips Big Tech Over AI Copyright

Runway Act 2 is Here | Full Facial and Gesture Motion Tracking

ChatGPT Agent | Autonomous AI Assistants

MultiTalk | AI videos with Audio, Motion and Reference Control

Matt Wolfe | Using Veo 3 Like a Pro

OpenAI's Internal Model Almost Wins World Coding Competition

How to make AI Music Videos for SUNO AI Songs in just 1 Click

Fireship | AWS just released its Cursor Killer