LM junior

@takuz0_ Perfect, but I'll soon have the final part to connect the DGX units to each other as well—though I'm not sure yet how I'll go about it. I'll work with my agents to help me.🫠

Testing 8x DGX Spark nodes in an open world setup :)

@LD_Theron mikrotik.com/product/crs804… can connect 4, 8 or 16 depending on the breakout cable being used. 400to2x200 or 400to4x100

Kimi K3 Coder Reap on 8xGB10 - 32t/s avg gen speed in coding task. This is while waiting for the full 16xgb10 setup to run the full k3. Considering that this one has same no of active experts, it is a good speed for my first, unoptimized, run of KIMI-K3-CODER-REAP-320-MXFP4 with dspark. I will work on improving these results and let you know.

🚨Grok 4.6 expected next week: • Launches around August 7th • Still 1.5T parameters • Significantly improved SFT and RL, almost like a perfected Grok 4.5 is how I see it • Grok 4.7 (2.1T) to drop a few weeks later too The cadence is absolutely ridiculous, they could actually catch the frontier imo, they have the infrastructure, talent and also training data not many people can match currently. What are your opinions on Grok 4.5 so far also?

@unseenmars_ OK , Thank you, I knew there was still stock available.

@LD_Theron we ordered them from amazon🖖

iykyk

AI takes less time to process and responds to my queries. Several papers on improving inference, speculative decoding, and many other topics have been published recently. Thank you all for your work.

I don't know what happened with Hermes agent, I was only away for 2 days and I came back to change everything, I don't know if it's because I was on wsl and then came to the powershell version but I'm won over. Thanks @NousResearch

China makes excellent models, but the problem that deserves attention is this one

Just got home from an amazing trip with the team to @nvidia HQ for the last couple days! Jensen even signed one of our team members' new Spark. Thanks @JensenHuang for having us!

Real !!

LOL 😂

It's not about having an AI like Grok, but about having an AI that performs well for the desired tasks, because I don't think anyone would want an entire cloud at home😭

Hermes Agent is so good, I quickly adopted it and use it daily. It's excellent for multitasking, and it's also good for coding specifically, but one person can't do it all. Perhaps they'll release a version specifically for coding. Otherwise, it's great.

We got attacked by secret unreleased proprietary models and defended ourselves with an open model, more precisely the @nvidia quantized version of GLM 5.2 coming from @Zai_org. Banning any open model would hurt first cyber security defenders, startups, small companies, researchers and everyone who's not a frontier lab and need on-prem affordable controlable models to compete and protect themselves. Let's not do that!

Kimi K3.1 is confirmed and targeting August. Mythos-level capabilities, open-weight. Faster inference, lower latency, more reliable coding and agentic workflows Expected to significantly close the gap with Claude Fable 5 The frontier open-weight race is moving faster than anyone predicted.

One of my favorite tools

Amazing !!

impressive !

@MiaAI_lab Okay

@junniorLB noup

Kimi K3 1-bit weights 594GB 🤯