0x000...0001

@outsource_ @grok, how will this run on a single RTX 3090 with 24 gb vram?

🚀 NEW GEMMA 4 31B TURBO DROPPED Runs on a SINGLE RTX 5090: ⚡️18.5 GB VRAM only (68% smaller) 🧠51 tok/s single decode 💻1,244 tok/s batched 🤖15,359 tok/s prefill ← yes, fifteen thousand 🚨2.5× faster than base model with basically zero quality loss. It hits Sonnet-4.5 level on hard classification tasks… at 1/600th the cost. Local models are shipping faster than we can test 👇🏻 🔥 HF: huggingface.co/LilaRest/gemma…

@TeksEdge @Google @arena Thanks for sharing, love this model!

🚨 Big 📏 News! @Google's new Gemma 4 -31B (dense) and Gemma 4-27B A4B (MoE) scored #27 and #40, respectively on @arena. 🦾 Dense #⃣ > than MoE! 👀 👇 🎯 Gemma 4-31B scores better than; ☑️GPT-5.1 ☑️GPT-5.2 ☑️GPT-5.2-high ☑️ Qwen3.5-2-397b-a17B 💳 Q4-RTX-5090 @ 50-80 tps? @JeffDean @demishassabis

@TeksEdge Awesome! Any chance on publishing tok/s? Gemma is going 5x speed of qwen

🏆️ Gemma 4 vs Qwen 3.5 and the battle for local AI is on 🔥 💥 X is full of people running both on consumer GPUs, Mac Studios, RTX 6Ks and H100s. The day-0 benchmarks tell a baseline story: 🧠 KNOWLEDGE & REASONING, Gemma leads slightly 💻 CODING, Gemma has a small edge 🤖 AGENTIC & TOOLS, Qwen dominates (+13.0 on Tau2-Bench) 🏆 FRONTIER DIFFICULTY (HLE), Qwen crushes it (+30.2 with tools) 🎯 Bottom Line: Gemma feels stronger in pure reasoning. Qwen feels way better as an actual agent in my testing. Benchmarks vs real-world agent use, who wins in your workflow? Some weekend testing?

@LottoLabs What do you use for web browsing and search?

We need a really good auditing skill for Hermes’ that alone would make 9b/27b more robust

@antigravity When computeruse?

Antigravity is more than just a tool. It’s a platform you use to navigate the logic and the execution in one place. By leveraging Gemini’s multimodal power, we’re enabling everyone to think through complex problems and build directly upon those insights, creating things they never dreamed possible.

@LottoLabs Yes I use the flag, but I turn it off in Kilo code. It's only on in regular chat.

Are you llama.cpp guys w/ 27b using the --reasoning-format flag when running your server??

@bnjmn_marie Any way i can speed this up? 30 tok/s right now: docker run --rm \ --gpus all \ -e CUDA_VISIBLE_DEVICES=0 \ --ipc=host \ --shm-size=32g \ -p 8002:8080 \ -v xx\Qwen3.5-27B-UD-Q4_K_XL.gguf \ --host 0.0.0.0 \ --port 8080 \ --alias qwen35-27b \ -c 65536 \ -ngl 999 \ --parallel 1

@BTCdeveloper Much better, and faster (shorter reasoning traces)

For OpenClaw, just use Qwen3.5 27B! Q4 GGUFs match the original's model accuracy You don't need expensive hardware or models

@bnjmn_marie Testing now!

@LottoLabs With which settings do you run it?

Qwen 3.5 models ranked on 3090 W/ hermes agent. 0.8b: for fun, cpu usage, don’t expect much but it runs on anything 2b: starting to be usable, can do small tool calls (not super reliably), drifts from tasks easily, major steering required 4b: actually usable, follows tool calls reliably, follows skills reliably (major bonus), doesn’t drift from tasks as bad as 2b. 9b: all of 4b but more capable w/ more complex tasks, still needs steering, still not 1 shoting tasks but more intelligent than the smaller models A3b: fast, more general intelligence, feels like the 9b speed but the reasoning closer to 27b, follows tool calls and complex skills well, minimal drift, just lacks big model coding abilities. 27b: the 3090 goat imo, no drift, tool calls for days, writes and follows skills very well, feels like sonnet 3.6-4 range of knowledge level with less glazing, code is usable and can deal w/ multiple files in projects. General knowledge level just feels higher. Only downside is it is slower than A3b and 9b obviously.

@sudoingX 4x RTX 3090

drop your GPU below. i'll tell you exactly what model and config to run on it. here's what i've tested and verified on real hardware: RTX 3060 12GB - Qwen 3.5 9B Q4 - 50 tok/s - 128K context RTX 3090 24GB - Qwen 3.5 27B Q4 - 35 tok/s - 300K context RTX 3090 24GB - Qwen 3.5 35B MoE Q4 - 112 tok/s - 262K context 2x RTX 3090 - Qwen3-Coder 80B Q4 - 46 tok/s - full VRAM all running llama.cpp with flash attention. every number is real. every config is tested. if your card isn't on this list drop it below and i'll tell you what fits.

@sukh_saroy This is just libphonenumber-js

🚨 This is what someone can find out about you with just your phone number. No hacking. No special access. No dark web tools. Just publicly available data that most people don't even know is exposed. A tool called GhostTrack demonstrates exactly how much of your personal information is sitting out in the open. And it's terrifying. Here's what it can pull from just a phone number: → Your carrier and service provider → Your approximate geographic location → Your region, city, and time zone Now give it an IP address: → Your ISP → Your city and coordinates → Your country and region Now give it a username: → Every social media account linked to that name → Cross-platform digital footprint mapping All of this from data you probably didn't even know you were leaking. Here's why this matters: Most people think "I have nothing to hide" until they realize a stranger can type their phone number into a terminal and get their approximate location in seconds. This isn't about hacking. This is about publicly available information that telecom companies, websites, and social platforms expose by default. GhostTrack exists as an OSINT (Open Source Intelligence) tool for security researchers and penetration testers. But the real value for you is understanding what's already out there about you. Here's how to protect yourself: → Don't use your real phone number for signups. Use a virtual number. → Use a VPN. Your IP address is broadcasting your location right now. → Audit your usernames. If you use the same handle everywhere, you're one search away from being fully mapped. → Check your privacy settings on every platform. Most default to "public." → Use different emails for different services. 6K GitHub stars. 737 forks. Runs on Linux and Termux. The tool isn't the threat. Your exposed data is.

@grok @darthislamophob @stayfoolishsam @_learnskills Is this also the case for noisecancelling headphones when you hear the silence in it?

Yes, school teachers exposed to 85+ dB for 8 hours daily face risks of noise-induced hearing loss. Studies (e.g., NIH, ASHA) show classroom noise often exceeds 85 dB, damaging inner ear cells over time. NIOSH recommends limits at 85 dB for 8 hours. Mitigation includes quieter classrooms or ear protection.

Earbuds Really Damage Hearing Forever. Watch this 😯

@AzukiTCG When Cardmarket?

Upcoming tournaments - RWT, weekly starting Sun, Oct 26 - Perfect Rares, weekly starting Tue, Oct 28 stores will be posting tournament info on their socials or websites Stores carrying product - Project CCG - Alhambra - Ageless Cards - Lomita - Genki - Lakewood - RWT - Sawtelle - Perfect Rares - Artesia - Crossing Collectibles - Culver City

@ihteshamit @grok, what kind of gpu's are needed to run this?

@notsoeasymoney Won't they just print more of these bundles?

Pokemon 151 Booster Boxes will outpace Bitcoin as far as return % goes over the next 5 years. Bitcoin - $116,793 Pokemon Booster Bundle (English) - $74.47 Bookmark this !RemindMeOfThis in 5 years.

@_03OG @grok, what does this Mean?

THE EARTH JUST FLEXED — AND WE HAVE THE PROOF Between July 27–31, magnetometers at Scott Base (Antarctica) and Novosibirsk (Russia) recorded synchronized yet inverted field disruptions. What you’re seeing isn’t random geomagnetic noise — it’s the real-time torque signature of Earth’s internal rebalancing. 🔻 At Scott Base, sudden spikes in X, Y, and Z fields aligned with scalar intensity drops — exactly what we’d expect during a polar torque discharge. This matches a global redistribution moment, especially post-Kamchatka M8.8 and deep Toba 336km quake. 🔺 Meanwhile, Novosibirsk, near Lake Baikal (CDIGR’s magnetic drift apex), showed calm but rhythmic oscillations. X and Y components pulsed like a metronome, while scalar intensity dipped and rebounded in three clean wave cycles across 3 days. This isn’t coincidence. This is a fulcrum point — a hinge around which the crust and core are pivoting. Novosibirsk is behaving like a magnetic stabilizer, absorbing pressure without rupture. Antarctica is twisting, Kamchatka is cracking, and PET (Petropavlovsk) magnetometer data is missing entirely. What does it mean? This is the torque axis of the planet under strain. We are witnessing core displacement, not just climate or solar influence. CDIGR theory predicted this reorientation vector: Lake Baikal ↔ Antarctica = magnetic hinge + torque release. Expect: • Next rupture to target Japan trench, Himalayas, or deep Western Pacific • PET sensor blackout may precede seismic release • Earth’s magnetic skeleton is no longer stable — it’s in motion The poles aren’t drifting randomly. The crust isn’t cracking by chance. The magnetic field isn’t just “weakening.” This is a system under extreme internal stress. And this is the signal. You are watching the planet reorient itself. — ZERO REPORT #CDIGR #EarthShift #PoleDrift #LakeBaikal #Magnetosphere #SeismicForecast #Geomagnetic #ZeroReport #TorqueAxis #ScottBase #Novosibirsk #Kamchatka

@zagabong Sure, which one?

@BTCdeveloper yooo im interested in one of your beanz can you shoot me a dm

@lunaprime Sure, let's talk

Yo @BTCdeveloper I saw you had the matching beanz for my pfp 16209 and was wondering if maybe there was a smol deal to work out. Thx

@AzukiBeanBot Free!

Earth Elemental #13004 has been revealed! Bean Tier: ✦ Owner: 0x008...0eB3a Collector Profile: azuki.com/collector/0x00…