Jose Miguel

@Prakhar6200 I don't understand it either. After spending a month with a Battleimage GPU, the problem is the software and compatibility.

WOW 😳😱 Really Intel?

@sudoingX 16GB VRAM Arc B50 Pro 🥹

what VRAM are you working with? i'm planning my next round of benchmarks and i want to test what matters for YOUR hardware. drop your exact GPU below. model, quant, what you're running. i'll tell you if there's something better for your setup.

@sudoingX How can I get similar results using a Intel Arc B50 Pro with 16 GB VRAM? I tried without any result since qwen3.5-14B cannot quantisize the KV in this GPU :(

@TeksEdge @intel Will those of us who own an Arc b50 pro receive any benefits?

🚨 Big News From @Intel for Local / Sovereign AI on Arc Pro B70🔥 📟 Huge memory boost on Windows as up to 93% of system RAM can now be dynamically allocated to the 👉 iGPU 👈 (e.g. ~59GB on a 64GB system) 📈 @Intel just released its Q1 2026 Arc Pro Driver (v32.0.101.8515), and it’s important for affordable high-VRAM inference. What’s New: Official full support for the new Arc Pro B70 (32GB GDDR6) and B65 discrete GPUs 💎 Better support for built-in Arc Pro graphics in Core Ultra processors 💵 Price of the B70 Just $949, which is roughly 33% to 25% the cost of comparable NVIDIA RTX Pro 4000 / 5000-class cards with similar VRAM. 👀 Why This Matters So Much? 1. Competition for AMD and Nvidia 2. This is Intel finally getting serious about the local AI inference market 3. Until now, getting 32GB+ VRAM for local LLMs usually meant buying expensive NVIDIA cards (5090) or very pricey new pro cards The Arc Pro B70 brings 32GB VRAM at a consumer-friendly price, which holds bigger models such as 70B+ Q4 models, MoE models, and long-context agents on their own hardware without massive cloud bills or hunting for expensive used 3090s. Intel claims strong gains vs NVIDIA RTX Pro 4000 💰 Up to 2x tokens per dollar (🐧) 🪟 Up to 2.2x larger context windows (🐧) 🪙 Up to 85% higher token throughput (🐧) ⚡️ Up to 6.2x faster Time-to-First-Token (especially in multi-user/agent scenarios) (🐧) Does it work on Windows & Linux? ✅ Yes, both platforms supported Full WHQL driver for Windows 10/11 Linux support included (Ubuntu confirmed) with a strong oneAPI/OpenVINO ecosystem How does it speed up LLMs without CUDA? 1⃣ Intel doesn’t use CUDA. Instead, it uses its own stack 2⃣ oneAPI + OpenVINO (highly optimized inference toolkit) 3⃣ IPEX-LLM (Intel’s PyTorch extension for fast LLM acceleration) 4⃣ vLLM with XPU backend (native Intel GPU support) 📟 The card has 256 XMX engines (dedicated AI matrix units) that are very efficient at BF16 / INT8 quantized models. 📰 Bottom Line This driver + B70 launch is Intel’s clearest signal yet that they want a real slice of the exploding personal sovereign AI hardware market. For the first time, 32GB-class local inference is becoming genuinely affordable and accessible. Huge potential step forward for private, on-device AI. What do you think? Is this a legitimate change for local LLMs or still waiting on real benchmarks from @digitalix? 👇

@no_stp_on_snek Battleimage community also needs some love :(

Working on a diagnostic script to benchmark TurboQuant across different hardware. Already have solid coverage on Apple Silicon (M5 Max 128GB, M1 Max 64GB) from folks in the community, but need more variety, especially NVIDIA GPUs. If you’ve got a different setup and want to help test, reach out.

@sudoingX 🙏🏻🙏🏻🙏🏻

i hear you. been hearing this from DMs, replies, and threads for weeks. everyone running local models is stuck digging through tweets and reddit posts for configs that should be in one place. that's why i've been collecting data. the polls, the "drop your GPU" threads, every benchmark i've published. it's not random content. it's the foundation for something i'm building. a place where you search by GPU or model and get verified configs, tok/s, VRAM usage and exact flags. community submissions. real numbers from real hardware. no influencer guesses. no spec sheet theory. tested data you can copy-paste and run. i'm close. more data comes in every day. more people are contributing. if you have a 3090 and can't get it running well, that's exactly the problem this solves. patience. it's coming.

@sudoingX Intel battleimage is in play?

if you're planning a new build or figuring out your architecture, post it in the community. we'll help you get the most out of your hardware. the goal is to help each other own our compute and stop feeding prompts to corporations mining your cognition.

i just became a mod of x/LocalLLaMA. if you're running local models on your own hardware and want in, the community is open. pinned and highlighted on my profile. approving members starting today. drop your setup below and i'll get you in. 3060, 3090, 4090, 5090, AMD, whatever you're running. all welcome. if you're hitting issues with hermes agent, llama.cpp, model selection, configs, i'm here. let's make local AI accessible for everyone.

@sudoingX In fact, I'm having trouble finding a support community for OpenVino with an Arc Pro B50. Updates to the latest models for this GPU are very slow ☹️

hey if you're running hermes agent on a 3060 or any single GPU and hitting issues, drop them below. i've tested on this exact card and i'll help you get it running. setup problems, config issues, model selection, optimization. all welcome.

@HaihaoShen @Alibaba_Qwen Thank you. Please, any ETA for the OPenVino format?

🚩More Qwen3.5 INT4 SLMs are now available - smaller yet strong. @Alibaba_Qwen 🫡🍺 huggingface.co/Intel/Qwen3.5-… huggingface.co/Intel/Qwen3.5-… huggingface.co/Intel/Qwen3.5-… huggingface.co/Intel/Qwen3.5-… huggingface.co/Intel/Qwen3.5-…

@SantaTeclaSV Tendrán una segunda edición de este diplomado?

¡Inscríbete a la capacitación de oratoria! Aprenderás a hablar con naturalidad y seguridad, controlar tus nervios, lenguaje corporal, estructurar tus ideas con claridad, mejorar tu tono de voz y dicción. 📍 Café Palacio. ✅ $25 (desayuno incluido). Inscripción: 7747-6100.

@TeamLignite It is really a awesome watchface. Congrats!

We will offer the option to hide the shadow in the next update 😉

We are so back. Our initial collection of eight beautiful watchfaces for @Pebble is now live! 🎉 Check them out: Lignite.io/pebble (We have many more designs coming up, stay tuned! ❤️) @ericmigi #pebble #watchface

@TeamLignite @Pebble @ericmigi Dashboard watchface is solo cool! Just one suggestion: change the time shadow to a lighter gray tone.

@VisualStudio It does not work for me: {"error":{"message":"Client specified an invalid argument","code":"invalid_request_body"}} Visual Studio Community 2022, v17.14

Big news for developers! Grok Code Fast 1 is now available in Visual Studio. This advanced AI model brings smarter, faster coding assistance right into your favorite IDEs via GitHub Copilot Chat. Available in public preview for Copilot Pro, Pro+, Business, and Enterprise plans, Grok Code Fast 1 supports agent, ask, and edit modes, making coding more intuitive and efficient. 👉 Admins: Enable the Grok Code Fast 1 policy in Copilot settings to unlock access for your team. 🔗 Read the full changelog to learn more and get started: msft.it/6016sIhco #AIcoding #GitHubCopilot

@ericmigi LCD_DOT, but I think the original developer is needed to check about the weather info

post name/link to your fav watchface and I will try out some on PT2

Mindblowingly cool breakthrough - we made it so all existing Pebble watchfaces/apps now automatically scale up to fill larger and higher resolution Pebble Time 2 display!

@its_hard_2_name @Pebble @ericmigi I'm having the time of my life tinkering with my pebble time again

It‘s been a long time... @Pebble @ericmigi

@GOGcom @TeamCherryGames Thank your @GOGcom ! at $6.99!! There is no other better store to buy Silksong!

HOLLOW KNIGHT: SILKSONG IS OUT NOW!!! 👉 bit.ly/Silksong_GOG There’s nothing more that needs to be said. Grab this masterpiece DRM-free on GOG and keep it yours forever. The whole gaming world has been waiting for this day, and it’s finally here 💜 @TeamCherryGames

Y entonces, concluimos que también somos una inteligencia artificial?

I made a YouTube channel for our community and farm ALL in Spanish here in beautiful El Salvador. If you’re looking to support, please give it a follow youtu.be/LPcvz--TSOg

@njelsalvador @rivenalvarez @MarkOfBitcoin Practical approach to massive adoption: let's pay public transport in SATs. Let the people use Bitcoin as any other currency. Normal citizens don't keep assets, just live day by day.

@rivenalvarez @MarkOfBitcoin Great question. And challenging to answer. There is a general lack of knowledge around what money is. People are generally lazy. People are generally skeptical, ironically often more of things that are good. People often need pain to change. Lead by example is a great start.

I can confirm that Berlin is in fact the most Bitcoin-friendly, widely-adopted circular economy in El Salvador, and I would guess the world. When you can buy a Papusa and have a Bitcoin conversation with a street vendor at 3pm on a weekday, THAT is adoption.

Recibi esta comunicado de mi Proveedor de Internet y parece legitimo :( No me quejo del servicio, la verdad ha funcionado bien, pero parece coincidir con el aumento al salario minimo y no se vale. Mejor incrementen sus ingresos vendiendo mas, lo cual creo que si ha sucedido.