Gloom
24 posts
Gloom
@Xyoskeleton
research @CORSAIR | local compute is fun :)
SF Beigetreten Ocak 2014
17 Folgt234 Follower
@Xyoskeleton Exactly, the way I view Intel now is how I viewed rocm even a year ago
the bmg cards are underutilized now but they have so much potential to raise the ceiling and I'm hoping to contribute to that
English
I almost bought a dgx spark at $2999, then I almost bought a rtx pro 5000 for $3400. Watched as prices continued to march.
Not getting left behind, I decided to throw my hat in the ring with Intel, got 2xB70s for $1900
Has barely been 2 days and I can't believe how much fun this is. Excited to contribute to Intel optimizations and provide useful AI to my family and friends
English
@Xyoskeleton @icanvardar Agree from my experience. And “normal people” is anyone that doesn’t work in tech, which is the majority.
English
google already lost the llm race
no one i know uses gemini anymore, i don’t even hear it mentioned that often
English
Given the vibe on X, prices of the GB10 are set to go up. Time to hoard? Also, who at Newegg thought they needed to entice buyers with a free mouse?
English
@BLUECOW009 Indeed not consumer viable yet, will get there maybe around 2028, I think we’ll see even cheaper api pricing for good local models I.e GLM before we see consumer hardware get cheap enough and accessible for most people.
English
@JoelDeTeves Depends on your use case. If you are going to buy at least 2 sure, if you buy one and not sure what kind of workflows you’re looking to experiment with you’ll probably end up regretting it tbh. Just stick with your dual 3090 setup imo and save up for a bigger upgrade.
English
Are DGX Sparks worth owning at this point?
The limited memory bandwidth doesn't seem practical for production workloads
English
@asayanmax How was it building in the iCUE link system ? Was it easier to manage ?
English
@Xyoskeleton @NVIDIAAI feel free to use it however you want, just tag me 😁
English
I fine-tuned Gemma 4 12B with Fable-5 style reasoning and assistant traces and released it as Gemmable 4 12b.
**Available formats:**
- GGUF: q4_k_m up to fp16
- MLX (for Apple Silicon)
MTP drafts are also included in the repo.
Training done on my @NVIDIAAI DGX Spark.
huggingface.co/Mia-AiLab/Gemm…
English
It's fun testing the @AMD AI TOP cards @GIGABYTEUSA at the lab, I want to eventually test four of these 😭 excited for more driver maturity on ROCm
English
Introducing GLM-5.2: Frontier Intelligence, Open Weights
- Significant improvements in coding and agentic tasks
- Strong long-horizon capabilities with a 1M context window
- Two levels of reasoning effort: GLM-5.2 (max) pushes the limits, while GLM-5.2 (high) strikes a strong balance between performance and token efficiency
- MIT-licensed open weights
- Same API pricing as GLM-5.1
Tech Blog: z.ai/blog/glm-5.2
Weights: huggingface.co/zai-org/GLM-5.2
API: docs.z.ai/guides/llm/glm…
Coding Plan: z.ai/subscribe
Chat: chat.z.ai
English
@LottoLabs Ahahaha would love to show you around the office if you are ever in the bay!
English
@sharbel This is true, local models are good at scale where you have dedicated budget to build out the infra. until costs decrease for hardware this will continue to be a very strong friction point for people. Also it’s hard to beat just the ease of access and reliability.
English
Most people trying local models this week will be back on the cloud API the moment restrictions lift.
The ones who actually set up a local model are a small, motivated minority.
Hardware is still a major wall for most people. A cheap Mac Mini runs a 7B model but the experience is noticeably worse.
Until that gap closes, this stays a power-user thing.
Alex Finn@AlexFinn
I don't care what kind of hardware you have, you should be running local models Governments are now banning models. They’re determining what technology you can and can’t use With local models, you are free and nobody can control you Even if you're on the cheapest Mac Mini you can be doing this Here's a complete guide: 1. Download LMStudio 2. Go to your OpenClaw/Hermes and say what kind of hardware you have (computer and memory and storage) 3. Ask what's the best local model you can run on there (probably will be Gemma 4 or Qwen. if you have a big computer, it will be GLM) 4. Ask 'based on what you know about me, what workflows could this open model replace?' 5. Have OpenClaw walk you through downloading the model in LM Studio and setting up the API 6. Ask OpenClaw to start using the new API Boom you're good to go. You just saved money by using local models, have an AI model that is COMPLETELY private and secure on your own device, did something advanced that 99% of people have never done, and have entered the future. If you are on smaller hardware you probably are not going to replace all your AI calls with this, but you could replace smaller workflows which will still save you good money Own your intelligence.
English
@tunahorse21 Updating documentation is like the one thing you definitely want to do manually 😭😭 I’ve had interviewed post grads were they have literally just offshore their thinking/coding to agents it’s really disappointing
English
@catalinmpit Curious on what kind of tasks you are wanting from Local LLMs to achieve? we have pretty much swapped our entire stack at the lab and run services through local models. We still use frontier models for review and initial planning but the cost offset made this worth it for us
English
See what it says at the bottom of the sale page that sells a similarly spec’d AI Max+ 395 PC w/2TB?
David Hendrickson@TeksEdge
🚀 AMD Ryzen AI Halo is now available for pre-order! A compact local AI developer platform powered by the Ryzen AI Max+ 395: 🧠 128GB unified LPDDR5x memory ⚡ 40 CU Radeon 8060S graphics (RDNA 3.5) 📦 Run models up to 200B parameters locally 🖥️ Windows + Linux support out of the box Build and deploy AI workflows without cloud dependency. Pre-order → @ amd
English
@LottoLabs Yes actually! Q3/Q4 release, new RDNA generation more memory and bandwidth should be really promising and should be competing with SPARK and potentially doing better in other areas. The question I have mainly is will ROCm support get better by the time of release.
English
