Glows.ai

Excited to support @deepseek_ai DeepSeek-V4 on Glows.ai 🚀 DeepSeek-V4-Flash is fully preloaded and auto-starts (API ready at: 8000). For DeepSeek-V4-Pro, we offer a VLLM-based environment, pre-configured for instant deployment. Spin up in minutes — no setup needed. 👉 Just launch and build. glows.ai

@ModernGrindTech Yeah we’re using Qwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled-GGUF/Qwen3.5-27B.Q8_0.gguf. So pretty heavy (~27B), which likely makes the gap more obvious. Env’s in the screenshot if you want to try it.

@glowsai solid numbers, the L40s util jump makes sense for steady inference but that 4090 cold start is brutal. Whats the model size youre benching? curious if the gap narrows on smaller weights or if its ha

We just launched new AI environments on Glows.ai: — Qwen3.5-27B (Claude Opus distilled, strong reasoning on 1 GPU) — Gemma 4 31B — SGLang 0.5.10 From 4090 → H100, all ready to run. No setup. No infra pain. Built for agents, LLM apps, and coding.

Hi David, cold-start time, inference speed, and GPU utilization during inference vary quite a bit by GPU: 4090: ~35s load, ~3.5 t/s, ~30% util L40s: ~7s load, ~23.5 t/s, ~98% util This difference is likely because of model + hardware efficiency. We usually recommend L40s for this model, it makes much better use of the GPU and gives significantly higher throughput. On Glows.ai, this model runs with llama.cpp: WebUI + API both supported, We’ll be publishing a full setup guide later this week.

@glowsai qwen distilled from opus on a single GPU is the spicy one. been wanting to test a local fallback for my voice-composer that doesn't pager me when anthropic has a bad afternoon. how's cold-start time on the 4090 env?

Most AI teams don't have a compute problem. They have a waiting problem. Waiting for the instance to spin up. Waiting for the dataset to download. Waiting for the environment to configure. By the time the GPU is actually doing work, hours have passed. So we built Datadrive to solve the part nobody was fixing. Teams were spending 30–40% of their GPU time just moving data around — downloading the same datasets, unpacking the same archives, over and over. That's not a training problem. That's a plumbing problem. Datadrive keeps your datasets hot and ready, so when your cluster spins up, your data is already there. No re-downloading. No unpacking. Just the work. If your GPU bills are high but your actual training time feels low - you’re not buying compute. You’re buying wait. glows.ai/datadrive

AI is moving from chatbots → autonomous agents 🤖 Projects like OpenClaw show what happens when AI can actually do things — run tools, automate workflows, and interact across apps. 🚀 We wrote a quick guide on how to get OpenClaw running quickly on Glows.ai. Read more 👀 glows.ai/article/opencl…

Glows.ai wishes you a prosperous and joyful Year of the Horse! 🐴

Are you part of Taiwan’s AI industry? Or maybe still on the way, hustling to get closer? We’re hosting a casual AI industry meetup — no slides, no panels, just real conversations over drinks. A chance to meet new faces, share ideas, and discover opportunities together. 🍷 Glows.ai AI Industry Meetup 🗓 Thu, Sept 18 · 18:30–20:30 📍 Taipei · A Glass or Two ✨ Limited to 25 seats ｜ NT$500 minimum (includes food & drinks) 👉 Sign up here: luma.com/7kkg9b1u

📢CPU instances now available, with the lowest configuration of 4C8G at just $0.18/hour! Start building now with our complete selection of CPU instances → glows.ai

How to Create a Realistic Voice with AI? A Live Demo of the BreezyVoice Model on Glows.ai Curious about making AI speak in your own voice? From recording, building speaker embeddings, to generating synthetic speech — this hands-on workshop will walk you through the entire process. We’ll demonstrate how to use the BreezyVoice model on Glows.ai to quickly create a personalized voice clone! 🗓️ Date & Time: July 4 (Friday), 8:00 – 9:00 PM 🔗 Google Meet: meet.google.com/opj-oood-upm

📢 【Event Recap】2025 Enterprise Generative AI Application Forum We were honored to collaborate with our strategic partners, TPIsoftware and Chief Telecom. Together, we showcased how conversational AI and private deployment can drive intelligent, real-world knowledge management solutions for the enterprise. A heartfelt thank-you to all our distinguished guests for your participation and support. Looking ahead, we remain committed to working closely with our partners to empower enterprises in harnessing the full potential of generative AI for digital transformation.

🆕 New GPU Launch: L40s Now Available! Build Your AI Creation Engine Glows.ai has just added a new GPU model! Introducing the NVIDIA L40s, designed for AI inference, image generation, and machine learning workloads — delivering powerful performance and stable compute. Ideal for: Text-to-image generation (Diffusion / SDXL) Large language model inference Multitask AI applications 📍 Now available for rental — come experience the power of this AI workhorse! 🔗 Start deploying: glows.ai

📢 RTX 4090 Pricing Adjustment: Now $0.49 USD/hr on Glows.ai Thank you for your continued support of our services. The three-month promotional offer for RTX 4090 instances has officially ended. To continue delivering value to our users, we’re introducing a new pricing plan effective June 13 (Friday) — RTX 4090 will be available at a special rate of $0.49 USD/hour. Start deploying now and take advantage of our full GPU lineup → glows.ai

✨How Should Enterprises Cross the Last Mile of AI Adoption? Tired of scattered documentation, overloaded knowledge base system, or fragmented customer service? It’s time for a generative AI solution that understands enterprise workflows. 🎯 On June 19, join us for an exclusive in-person seminar where 3 industry leaders will share their strategy for effective AI implementation: ✅ 昕力資訊—Developer of the SysTalk.ViKi platform Taiwan’s leading developer of generative AI conversation tech, empowering enterprises to build smart Q&A systems, AI-powered knowledge management, and customer service integration. ✅ KonstTech – Cloud Deployment Experts Specialized in high-performance GPU cloud infrastructure and private AI training environments. In Partnership with Glows.ai, deliver stable computing capacity to universities and enterprises. ✅ Chief Telecom Inc. 是方電訊 – Enterprise Security & Hybrid Cloud Specialists Trusted leaders in IDC infrastructure and cybersecurity. The backbone for secure, large-scale AI deployment. 📌 [2025 Enterprise Generative AI Implementation Forum] 🗓 June 19 (Thu)｜13:30–16:30｜Taipei 🎟 Free Registration: lu.ma/dct7i3fe 👥 Who should attend: IT/Ops Managers, AI Implementation Teams, System Integrators, and Tech Consultants 🎁 Sign up for presentation decks, tailored solution recommendations, and a chance for personalized consultation!

Analyze Market News with FinGPT and Rapidly Build a Trading Bot on Glows.ai! Ever wondered if AI could read market news, detect sentiment, and even make trading decisions for you? In this workshop, we’ll build a FinGPT-powered trading bot: 🧠 Leverage the FinGPT model from Hugging Face 📰 Fetch real-time market news and perform sentiment analysis ⚙️ Deploy in one click on Glows.ai to create your own AI-driven trading prototype 🗓 Date: June 6th, 15:00 (GMT+8) 🔗 Join us: meet.google.com/opg-xjka-ehn

At NCKU, we had the pleasure of joining a class at the Institute of Manufacturing Information and Systems to introduce Glows.ai! Through a hands-on session led by our instructor, students explored how to deploy models directly on the platform, including a live demo and the chance to run their own RAG pipeline in real time. Huge thanks to NCKU for the invitation, and to all the students who participated and brought such great energy to the session!

Honored to take the stage at #COMPUTEX2025! Glows.ai was invited to present our global GPU infrastructure showcase. Our very own sales team member, Charisse, took over the spotlight to share how Glows.ai helps teams move from raw compute to ready-to-build AI environments in just minutes — with zero DevOps, and full-stack flexibility. From instant deployment to intelligent scaling, Glows.ai is reshaping how developers interact with AI infrastructure. Thanks to COMPUTEX 2025 for the stage and the opportunity to share our mission with the community and investors!

🆕 New Image: BreezyVoice WebUI is Live! 🎙 Feature Highlights: Developed by MediaTek Research, BreezyVoice is an AI voice synthesis system designed specifically for Taiwanese Mandarin. It supports Zhuyin (Bopomofo) input and voice cloning — generating smooth, natural speech from just a 5-second audio sample. Perfect for AI dubbing, character voice creation, and other creative applications. 📦 Specs: RTX 4090 x1 🎮 Try it now: glows.ai

🚀 COMPUTEX DAY 1 — Glows.ai is Live! We're live at #COMPUTEX2025! Come check out how Glows.ai + KONST are transforming AI deployment — from local GPUs to one-click cloud-based inference and training. 💡 Meet our team, watch live demos, and get hands-on with Glows.ai. 🎁 Bonus: Visit us to unlock free cloud credits. 📍 Booth M1333a, 4F, Nangang Exhibition Center, Hall 1 🗓 May 20–23 | 🕘 09:00–17:30