Kimi K2 Technical Breakdown: How It Challenged AI’s 7-Year Status Quo
Related videos
Trader Joe's vs. Aldi Cooking Challenge
Mythical Kitchen
222.0k views
A Year Into Making LLMs, and now Topped Open Source SoTA?!
bycloud
22.1k views
BREAKING: Trump Suffers MENTAL BREAKDOWN as IT ALL CRUMBLES!
Luke Beasley
131.1k views
BITCOIN: ANOTHER LEG DOWN STARTING!!! (how I profit from the bear)
Ivan on Tech
15.5k views
How DeepSeek V4 Broke AI’s Cost Curse
bycloud
101.8k views
China CHALLENGES Trump's Iran Blockade
Breaking Points
535.7k views
This is Why You Never Challenge a Navy SEAL…
Shawn Ryan Show
347.8k views
The Chinese AI Iceberg
bycloud
107.4k views
The View Has a Collective Nervous Breakdown Over RFK Jr
Kim Iversen
32.9k views
I challenged Manus AI to find $1M+ AI Startup Ideas - It Actually Delivered
Greg Isenberg
198.2k views
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Related videos
Trader Joe's vs. Aldi Cooking Challenge
Mythical Kitchen
222.0k views
A Year Into Making LLMs, and now Topped Open Source SoTA?!
bycloud
22.1k views
BREAKING: Trump Suffers MENTAL BREAKDOWN as IT ALL CRUMBLES!
Luke Beasley
131.1k views
BITCOIN: ANOTHER LEG DOWN STARTING!!! (how I profit from the bear)
Ivan on Tech
15.5k views
How DeepSeek V4 Broke AI’s Cost Curse
bycloud
101.8k views
China CHALLENGES Trump's Iran Blockade
Breaking Points
535.7k views
This is Why You Never Challenge a Navy SEAL…
Shawn Ryan Show
347.8k views
The Chinese AI Iceberg
bycloud
107.4k views
The View Has a Collective Nervous Breakdown Over RFK Jr
Kim Iversen
32.9k views
I challenged Manus AI to find $1M+ AI Startup Ideas - It Actually Delivered
Greg Isenberg
198.2k views
Top Comments (10)
You got some comments in your bots 💀. Anyways, I’ve been using Kimi for a bit through t3 and it’s been relatively decent! Especially for such a low cost model (Super great explanations with the Minecraft reference. What a great way to explain local optimization!)
I checked your channel earlier today and was like: Huh, guess you won’t be able to upload for a while. Glad I‘m wrong.
Another great video! btw hope mandatory military training is treating you well
When they say ‘one expert per GPU’, my understanding after looking into it is that it's about the training, not inference. In serving, multiple experts live on each GPU, so you don’t actually need 384 GPUs to run a single instance. After watching again I realized you also said that "...to train", but I thought it was an interesting thing to highlight anyway. I was not aware of the huge difference in requirements between training and inference before. Hope you are enjoying your military duty :D
Kimi k2 passed my vibe check. Many good MoEs give it unprecedented depth I haven't seen before.
Master AI agents now using HubSpot's FREE resource! https://clickhubspot.com/aa716f
this model is nuts......using it 3 days non stop.
Great breakdown. Another very interesting thing to point out is their synthetic data generation strategy, especially with fake MCP tools to make it a better tool calling LLM.
Oh wow, way more excited about the new muon optimizer than anything LLM related
😊nice video. time to try muon in the next project.
Unlock the Data Inside
Turn Videos into Knowledge
- Get FREE 10/day: transcripts, summaries, chats
- Chat with videos, export text & PDF
- $1 free API credit for RAG, chatbots & research
Free forever plan • All features unlocked
Top Comments (10)
You got some comments in your bots 💀. Anyways, I’ve been using Kimi for a bit through t3 and it’s been relatively decent! Especially for such a low cost model (Super great explanations with the Minecraft reference. What a great way to explain local optimization!)
I checked your channel earlier today and was like: Huh, guess you won’t be able to upload for a while. Glad I‘m wrong.
Another great video! btw hope mandatory military training is treating you well
When they say ‘one expert per GPU’, my understanding after looking into it is that it's about the training, not inference. In serving, multiple experts live on each GPU, so you don’t actually need 384 GPUs to run a single instance. After watching again I realized you also said that "...to train", but I thought it was an interesting thing to highlight anyway. I was not aware of the huge difference in requirements between training and inference before. Hope you are enjoying your military duty :D
Kimi k2 passed my vibe check. Many good MoEs give it unprecedented depth I haven't seen before.
Master AI agents now using HubSpot's FREE resource! https://clickhubspot.com/aa716f
this model is nuts......using it 3 days non stop.
Great breakdown. Another very interesting thing to point out is their synthetic data generation strategy, especially with fake MCP tools to make it a better tool calling LLM.
Oh wow, way more excited about the new muon optimizer than anything LLM related
😊nice video. time to try muon in the next project.