正在加载视频...

视频加载失败

Easter loading… but the data is already here. Start your weekend early with #GagaWeekend: UGX 6,000 4.5GB + 20 MTN mins UGX 8,000 6GB + 20 MTN mins UGX 10,000 8GB + 20 MTN mins Dial *100*0# or use MyMTN App.

14,506 次观看 • 3 个月前 •via X (Twitter)

0 条评论

暂无评论

原始帖子的评论将显示在这里

相关视频

Run Gemma 4 26B MoE on 8GB VRAM with 250k context at 20+ tokens/sec If you own any 8GB VRAM graphics card, stop what you are doing. Local AI just had its absolute "Holy Shit" moment for budget hardware. Yesterday, I benchmarked Unsloth Gemma 4 12B Q4_K_XL on an 8GB card. The community went wild but immediately demanded more: "Can we run a 25B+ model on budget GPUs?" Today, I’m delivering exactly that. I am running a massive 26B parameter Mixture of Experts (MoE) model locally on a standard 8GB VRAM setup with 250k full native context!. If you own an RTX 3060, 3070, 4060, or any budget GPU with 8GB of VRAM, the local AI paradigm has completely changed. The performance metrics are astonishing: - 20 tokens/sec flat decode throughput. - Stable, flat decode speed even with massive prompts. - I threw a 60k token prompt at it, and it still clocked in at 20 TPS without dropping a single frame. # What about prefill? Yes, Time To First Token (TTFT) is slightly high when swallowing massive contexts. But with a solid 200 tokens/sec prefill speed, the wait is barely noticeable and highly usable. And this is running completely without Multi Token Prediction (MTP) active. How is this possible? It’s the magic of Google's new QAT (Quantization Aware Training) quants for Gemma 4. The model weight file (unsloth gemma-4-26B-A4B-it-qat-UD-Q4_K_XL.gguf) is only 13.2 GB, making it the ultimate local powerhouse. # The Test Setup: CPU: Intel Core i7 RAM: 16GB System RAM GPU: NVIDIA GeForce RTX 4060 Laptop GPU (8GB VRAM) # The Secret Sauce (The -cmoe Flag) To make this work properly on any 8GB card, you must use the -cmoe (CPU MoE) flag in llama.cpp. This flag isolates the heavy MoE expert weights directly to system memory (CPU/RAM) while letting your GPU focus strictly on the Attention layers and the KV Cache. It prevents VRAM spillage and holds the throughput rock solid. # The flags: -m "gemma-4-26B-A4B-it-qat-UD-Q4_K_XL.gguf" -cmoe -c 248000 -v Once running, just open the UI on localhost and toggle the new reasoning lightbulb icon in the text input box to watch the model perform multi step thinking. Are you still running smaller models, or are you ready to scale up your budget local setups? Let's discuss in the replies

Alok

292,032 次观看 • 1 个月前

you're paying $20/mo for something your $500 GPU can already do. Gemma 4 26B A4B QAT MoE + Hermes Agent running on a single RTX 4060 (8GB VRAM). Built a vision capable, 100% free, 100% local, private AI assistant that lives in my Chrome browser. No API keys. No cloud. No subscriptions. 100% vibe coded. 0% handholding. It has full context of whatever's on my screen can answer questions, summarize pages, extract data, and see images. Same local model handles everything, no external calls, ever. keep reading for the model and hermes agent tips i learnt while building this locally. Here's the exact setup for anyone running local LLMs on 6-8 GB VRAM: llama.cpp server flags (on my NVIDIA RTX 4060 8gb VRAM): -m gemma-4-26B-A4B-it-qat-UD-Q4_K_XL.gguf --cache-type-k q8_0 --cache-type-v q8_0 -c 150000 --port 8080 Throughput with quantization: Prefill: 200-250 tokens/sec Decode: 20-25 tokens/sec reduce context if oom on 6 gb vram card. Key learnings: - Quantize KV cache to q8 for faster prefill/decode. Prefill goes from 100-150 (unquantized) to 200-250 tok/s (q8). - But watch out, once actual context grows past ~50k tokens on high entropy workloads, q8 KV quantization can cause hallucinations. Low entropy workloads are mostly unaffected. If you see it happening, drop the quantization. This is common across all local models. - In Hermes Agent settings -> Memory & Context, bump compression threshold from default 0.5 to 0.7. Default triggers way too frequent context compression and eats time. Up next: add persistent memory, web search, tool calling, streaming output and whatever you suggest. Running a 26B MoE with vision + 150k context window on 8GB VRAM would've sounded impossible 6 months ago. Works the same on the NVIDIA RTX 3060 Ti, 3070, 4060 Ti, 5060, 2080, or any 8GB card. VRAM is the only requirement. Local AI agents are closer than people think. You just need to know where the knobs are. Model's Unsloth quant hugging face link in the comments. Have you tried Hermes agent by Nous Research yet? What are you building with local LLMs? Drop it below, let's see what this community is shipping.

Alok

36,031 次观看 • 12 天前

6 months ago we were dropping a new app every week No one cared We’d randomly get 10k users on an app But they came for the app, not the person The thing is, I don’t care about making a retentive app for one audience I want to be a retentive person ~ for my audience What if I was the app? Not a single Paul Thomas Anderson movie is the same Different subject matter, different genres, so you’d assume different audiences However the same people that went to see his last film, came to see One Battle After Another So the demographic isn't dependent on the subject matter The demographic is just, Paul Thomas Anderson fans The software industry has long told that you need to work on one thing, for the rest of your life That’s not how art works tho, is it? Can you imagine telling Jay Z “Great job on the blueprint, now iterate on that same album for the next decade” The landscape of tech haas been stifling the growth of creators by not allowing them to explore other interests 6 months ago I said no to this "requirement", despite what everyone told me, and continued to drop what I liked every week The second a trend was happening on Tiktok, I had the app out that week Somehow 6 months later, the world is conforming to this ideology Instead of software creators limited to making apps for one audience and one niche, there’s a new world of ephemerality and expression What if instead of optimizing for users, we optimized for fans Making apps that are expressive of your life, your commentary, your heartbreak Garnering an audience that will follow you through each step of your story Each of those steps being its own app Why shoot for daily active users when you can get daily loving fans When fans use your app, it’s not just about resonating with the story, the app places them IN THEIR OWN story Here’s an example You’re a 20 year old girl who’s at UMiami You scroll through Tiktoks in your dorm room about “mogging”, a trend to outshine your friend in a photo You laugh and share videos seeing celebrities mog each other, but that’s the extent of it Then at Danger Testing we make an app called mog or not, where you and your friend can upload a photo and AI tells you who’s mogging Now you’re at the bar with your sorority sisters, playing all night, whether your winning or losing it’s the time of your life cause something is finally about YOU ENOUGH OF WATCHING MOVIES LET’S MAKE YOU THE MOVIE LET’S MAKE YOU THE STAR AN APPSTAR

los (appstar)

14,257 次观看 • 9 个月前

$25K+ profit daily from 1 wallet, with OpenClaw. I have the exact step-by-step guide, giving it free for 24 hours. To get it: 1. Comment "OpenClaw" 2. Like and Retweet. 3. Follow me Himanshu Kumar ( So, i can send you DM) I ran a simple script last night with Claude Code. Pull on-chain data from Polymarket, sort by win rate on 15 minute BTC markets. 20 minutes later, 100s of wallets showed up. Most were losing money or barely breaking even. Then I spotted 1 address. 200+ trades daily, every single week profitable, timing so precise it looked robotic. Because it is. I fed the wallet address back into Claude Code. Asked it to reverse engineer the strategy. 20 mins later the full breakdown appeared on my screen. Here is how it works: Bot monitors Binance and Bybit every 100ms. Waiting for BTC volatility compression to drop below 0.08%. When it hits that level, it buys both Up and Down contracts at 25 to 35 cents each. Classic straddle play. 1 contract loses, the other rockets to a dollar. Entry at 30 cents means 3x to 4x return every time. Repeats dozens of times per day. Result: $13K to $25K profit daily from 1 wallet. No human intuition, no insider tips. Just an algorithm exploiting a gap in market mechanics. I searched to see if anyone else found this wallet. Turns out yes. There is a Telegram bot that auto-copies trades from wallets like this. I connected it to the same address. Every entry matched what my terminal showed. You can now copy-trade an algorithm in real time. That capability did not exist 12 months ago. Comment "OpenClaw" and I will send you everything. Must Follow me Himanshu Kumar to get the DM.

Himanshu Kumar

13,086 次观看 • 4 个月前

Andrej Karpathy, the CEO of Obsidian, and Claude Code just built the smartest second brain on earth. It started with a 1-page gist that 21M people read. Karpathy frame flips everything you know about notes: Obsidian is the IDE, Claude Code is the programmer, and your notes are the codebase. You don’t ask AI questions it forgets by tomorrow you make it maintain a living wiki. 3 commands run the whole system. Ingest: drop an article, a podcast, a PDF, and Claude splits it into atomic pages linked to everything you already know. Query: ask anything and it answers from your own notes, in your own words, citing your own pages instead of guessing from training data. Lint: once a week Claude walks the entire vault, flags contradictions, kills stale claims, and wires orphan notes back in. Then Steph Ango made his move. The Obsidian CEO didn’t bolt an “Ask AI” button onto the app he shipped 5 skill files that teach Claude to write Obsidian’s native language: wikilinks, Canvas, Bases, the CLI. The repo crossed 13,900 stars in weeks and sits at 41,000 now. Karpathy runs it on his own reading: 100 articles and 400,000 words, cross-linked and maintained while he sleeps. No vector database, no embeddings, no $20 a month memory app just a folder of plain markdown and an agent that never gets tired of the boring part: the linking, the filing, the upkeep that killed every Zettelkasten since 1965. Your vault has 3,000 notes nobody will ever reopen. His read all of themselves by breakfast. Every app promised a second brain this is the first one that thinks.

West Lord

469,682 次观看 • 3 天前

Everyone is talking about @_kaitoai and its very rewarding Yaps Over 70% of CT is onboard already even smaller accounts If you are unsure of what Kaito Yaps are, Here’s a thread to break it down for you in 3 mins👇🧵 __________________________________________________________________________ What is Kaito about? Kaito AI is Web3’s Crypto Oracle where you can sniff out web3 info, golden projects, attention and earn rewards. Kaito is basically the hottest research hub for founders and users alike. It doesn't just stop there, Kaito's mechanism also makes it possible to earn rewards for your social traction Let's take a deeper dive on how it works👇 - Kaito finds alpha projects - You yap about them: Share smart takes on Kaito-approved gems. - Earn yaps! Good yaps = more treasure. Guide to Quality yaps👇 - Yap ONLY on Kaito’s picks. - Quality content only! Quality beats noise. Good quality yaps, more reward How to Climb to Yap Royalty🪙 - Hunt the hot trends: Focus on “trending projects” Kaito spots ‘em first. Stick to the buzz! - Copy the big guys: Peek at “leaderboard apes.” Learn their yaps, steal their moves (politely! And give shout out). - Timing is Key: Yap when people are more active If you miss out on Kaito, here’s what you lose: - Daily yaps(free money) up to 24, 000 yaps up for grabs daily Each yap could be worth a $100 - Early access to Alpha Projects🚀 - Attention. Top yappers get noticed, building reputation and audience. Connecting with whales, influencers and OG crypto hunters Imagine missing all that! Don’t be slow Get in early👏 Ready to smash the leaderboards? Use this link → Start yapping!

Defi Queen ( 💙, 🧡, 💜)

50,289 次观看 • 1 年前