We’re proud to sign the Open Weights and American AI Leadership letter.
AGI isn’t the work of one company. It’s built on generations of human knowledge and collective progress.
That’s why we believe AI should be open by default, for the benefit of everyone.
For my first post, I’m sharing a letter @NVIDIA signed on why open models matter.
AI will transform every industry, power every company, and be built by every country.
Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty.
The world needs both frontier closed models and frontier open models.
https://t.co/AUKzoQ5Ikb
Thank you, Singapore!
Last week, GMI Cloud and @tgbeings brought together around 30 C-level executives, founders, and senior leaders from the region's AI ecosystem for the AI Leaders Workshop.
Andy Chen, our Global VP of Business and Product Development, kicked off the evening with a talk on matching GPU compute to model needs in the AI-native era. Then a fireside chat with Sun Weiran from @OpenAI, Shivang Gupta from @tgbeings, focus on what it takes to move from frontier AI models to real-world agents.
The highlight was the executive roundtable. Four groups dug into choosing the right model, building a multi-model strategy, and scaling with control, then shared their key takeaways with the whole room.
We hope these conversations will stay with you. See you next time!
#GMICloud #AIInfrastructure #AILeadership #EnterpriseAI
We’ve secured $668M in funding, led by ARCHIV with @nvidia participating. 🚀
For builders in Singapore and across ASEAN, we’re expanding GPU capacity and scaling our inference platform to help bring more AI products into production.
#GMICloud#ASEANAI
📣 Reminder: our private workshop in Singapore is happening next week!
We're bringing together 30 C-level executives, founders, and senior leaders from the region's AI ecosystem for an afternoon of real talk: how GPT-6 is shifting the AI discourse, how to match GPU compute to your model's needs, and a roundtable on choosing the right model, building a multi-model strategy, and scaling with control over cost, security, and governance.
📅 Sep 29, 2026 (afternoon)
📍 OGBC Innovation Hub, Singapore
Co-hosted with The Generative Beings, Asia's largest AI founder and builder ecosystem.
Spots are limited and filling up.
If you haven't registered yet, now's the time 👉 https://t.co/VXfNuGPnuD
#GMICloud #AIInfrastructure #AILeadership #EnterpriseAI
🚀 GMI Clouders is landing in APAC
GMI Cloud's Ambassador Program has been growing builders and communities around the world — and now we're deepening our roots in Taiwan, Japan, South Korea, Singapore, and across ASEAN.
Two tracks. Pick one or both:
🎥 Content Creator: Build with GMI Cloud and share it. Demos, tutorials, reviews — on X, YouTube, or wherever your audience lives.
🤝 Community Builder: Host meetups, hackathons, and workshops in your city. We fund it, you run it.
What Clouders get:
✅ Monthly free GMI credits
✅ Event sponsorship for approved community activities
✅ Featured spotlight on GMI's ambassador pages
✅ Tiered referral rewards + exclusive swag
✅ Early access to GMI’s new products and events across APAC
Who we're looking for:
Developers, ML engineers, open-source contributors, and creators who love building and sharing — plus hackathon organizers, meetup hosts, and AI community leaders ready to grow their local scene.
👉 Apply now: https://t.co/dGrZNMWUru
We're heading to Tech Week Singapore, and we'll be at the booth with CloudEngine Digital!
Come meet the GMI Cloud team, see what we're building in GPU cloud and AI infrastructure, and chat with us about scaling your AI workloads.
📅 29-30 September 2026
📍 Sands Expo & Convention Centre, Marina Bay Sands
🏷️ Booth L5 5-126
Whether you're exploring GPU cloud solutions, scoping out your next AI project, or just want to say hi, we'd love to see you there.
See you in Singapore!
#TechWeekSingapore #GMICloud #CloudEngineDigital #AIInfrastructure
🇸🇬 On September 29, we’re hosting a closed-door C-level gathering in Singapore, bringing together executives, founders, and key decision-makers from across the AI ecosystem for an in-depth exchange.
Together with The Generative Beings, we’ll explore how GPT-6 is reshaping AI, how to match models with the right GPU infrastructure, and how to scale a multi-model strategy while balancing cost, security, and governance.
📅 September 29, 2026 | 2:00–5:30 PM SGT
📍 OGBC Innovation Hub, Singapore
Seats are limited and applications are subject to approval.
Apply to join 👇
https://t.co/VXfNuGPnuD
#GMICloud #AILeadership #AIInfrastructure #EnterpriseAI
🚀 FUTUREMODE 2026 wrapped after 3 incredible days!
Thanks to every developer, creator, founder, and AI builder who stopped by our booth to explore what's possible with AI 💪
As a sponsor of the BUILDMODE hackathon, we powered teams with GMI Cloud API Credits across 200+ AI models, and the projects that came out of it were amazing 🤩
Huge congrats to Exuan design Ai art for winning in the AI x Creativity Track with an AI short film built on GMI.
Our Front-End Technology Director, Fred Jhang, also hosted an AgentBox Launchpad workshop, where 90+ participants went from a single prompt to a fully deployed AI agent in just 40 minutes.
Thanks to @TaipeiWeek and SITCON for making this hackathon happen, and to everyone who came out to connect.
See you next time ✨
#FUTUREMODE2026 #GMICloud #AIInfrastructure #AgentBox #AgenticAI #Hackathon #MaaS
Huge thanks to Wan for teaming up with us on last weekend's AI filmmaking challenge in Taipei, bringing together 60+ creators across 19 teams to see what's actually possible with #AIvideo today 🎬✨
17 original short films made possible by Wan3.0.
Huge congrats to this year's winners 👏
Last weekend in Taipei, Wan and @gmi_cloud brought together over 60 AI video professionals across 19 teams for an immersive AI filmmaking challenge! 🎬✨
Powered by the latest Wan3.0 AI video models, participants created 17 stunning AI-generated short films within hours. We were also honored to welcome Golden Horse Award-winning directors who joined us for an insightful panel discussion and served as distinguished judges.
A huge congratulations to this year’s winners!
Winning videos: https://t.co/145ALnRREI
#AlibabaCloud #Wan #AIVideo
🎬 60+ creators. 19 teams. One surprise prompt. And just a few hours to turn ideas into films.
Using selected AI models on GMI Cloud MaaS, the teams created 17 original AI short films, each with its own unique story, style, and imagination. 🎥✨
And the winners are… 🥁
🥇 1st Place | Cocoon Room
🥈 2nd Place | Ma
🥉 3rd Place | I Love You So Much
🎨 Special Animation Award | Return Voyage
A huge thank you to our judges and panelist, Director Yang Li Chou , @donniepai , @wilson_pro_ai , Tess Xian from @Alibaba_Wan and Fred Jhang by @gmi_cloud , for bringing their creative, technical, and industry perspectives to the competition.
We also explored how AI is reshaping real film production through our creator panel and hands-on workshop.
Big thanks to @Alibaba_Wan ,@Alibaba_Qwen , SV Asia, and @Seagate for making this creative experiment possible.
AI brings the tools. Creators bring the magic. We can’t wait to see what they make next. 🎞️✨
#GMICloud #AIFilmmaking #AIVideo #AIGC #FutureOfFilm
📷What an incredible turnout for the GMI Cloud × Wan AI Video Professionals Competition!
Over August 29–30 in Taipei, we saw an amazing wave of creativity and passion from AI video creators pushing the boundaries of what's possible with generative video.
A huge thank you to our judges, acclaimed directors 楊力州 and 李念修, and to our workshop mentor 白家德, for lending their expertise and vision to this event.
To everyone who joined, submitted, and cheered along, thank you for making this such a special celebration of AI-powered storytelling.
See you next time! 🫶
Great to join the 2026 Japan–Taiwan Innovation Summit in Tokyo!
Tim Chen, CFO of GMI Cloud, joined Yushi Nagasaka (Kyocera), David Wacker (KONST), and Sam Ho (Rayleigh Vision Intelligence), moderated by Edgar Chiu (SparkLabs Taiwan), to discuss “Building Trusted AI Infrastructure: Opportunities for Taiwan–Japan Collaboration.”
We look forward to continuing to strengthen Taiwan–Japan collaboration and building the infrastructure that powers the next generation of AI.
Thank you Startup Island TAIWAN for bringing the two ecosystems together!
#AI #AIInfrastructure #GMICloud #JapanTaiwanInnovationSummit
Recapping a great conversation in Taipei last week 😎
Our CFO Tim Chen sat down with Marco Chung (@NYSE) and Victor Tseng (CFO, @kkdaytravel), moderated by Jennifer Lin (Headline Asia), to talk about what it actually takes for a fast-growing company to get ready for the US public markets.
Less about the IPO itself, more about building the discipline, systems, and governance that let a company scale without losing speed.
Thanks to Headline Asia for putting this together!
#GMICloud #AIInfrastructure #NYSE #CapitalMarkets
Meet "Jenny & Benny", a new character short series by @shirawiggles, brought to life with Seedance 2.5.
Can't wait to see where this duo goes next 😎
Seedance 2.5 is in GMI and I started this New Short Series
'Jenny & Benny'
She encourages him every single day. He only wants to take a nap every single time.
Check prompt on the replies 👇🏻
@gmi_cloud#gmicloud#gmiclouder#3danimation
New on GMI Cloud: MiniMax H3 🚀
It's the top-ranked open model in Video Arena, across both Text-to-Video AND Image-to-Video.
(Video created with @MiniMax_AI H3)
What would you generate first? A scene, a story, or something totally random?
Try it on GMI Cloud and drop your results below.
#GMICloud #MiniMax #AIVideo #TextToVideo #ImageToVideo
Big shift happening in AI right now.
LLMs have gotten seriously good at frontend generation, and open source models are starting to outperform closed source.
We put it to the test.
4 models, one prompt, and recreate the entire Apollo 11 mission as a cinematic 3D sequence in the browser.
2 open source (Qwen3.8-Max, Kimi K3) went up against 2 closed source (Fable 5, GPT 5.6 Sol).
Fable 5: the most faithful reconstruction, but came at 4x the cost. 1h30m · $40.5
Qwen3.8-Max: best cinematography of the four, hands down. 1h48m · $12.8
Kimi K3: slowest to finish, but rendered the launch better than anyone. 2h13m · $20.0
GPT 5.6 Sol: cheapest and fastest, but quality took the hit. 1h07m · $9.8
Open source is closing the gap fast.
#AI #GenAI #OpenSource #CloudComputing #AIInfrastructure
We're now the #1 provider for DeepSeek V4 Flash 0423 on @OpenRouter, with the lowest input price in the market at $0.038/M (about 40% cheaper than the #2 provider), and handling 24.6% of all daily tokens.
To everyone building with GMI Cloud, this one's for you.
Thank you. 🙌
#AI #GenAI #DeepSeek #CloudComputing #AIInfrastructure
Create a high-quality desktop 3D website called Retro Console Museum. Users can explore the museum in first person, inspect six detailed retro gaming machines, power them on, and play a fully functional original game on each machine’s physical screen.
Include these six exhibits: Vector One — a 1970s wood-and-metal home console with a monochrome CRT. Its game, Vector Rally, is a simple vector-style racing game controlled with a rotary dial.
Arcade Nova — a 1980s upright arcade cabinet with a joystick, buttons, coin slot, and color CRT. Its game, Neon Defender, is a side-scrolling pixel-art space shooter.
Turbo Disk 94 — a 1990s disc-based console with a wired controller and early 3D graphics. Its game, Polygon Drift, is a low-poly three-lap racing game.
Pocket Mono — a late-1980s monochrome handheld with a reflective LCD, cartridge slot, D-pad, and two buttons. Its game, Block Harbor, is an original falling-block puzzle game.
Crystal Pocket — a transparent late-1990s color handheld with visible internal circuitry. Its game, Forest Courier, is a top-down pixel adventure about delivering packages through a forest village.
Dual Flip — an early-2000s folding dual-screen handheld with a stylus. Its game, Signal Lab, is a circuit-routing puzzle played across two screens.
Every machine should have a unique startup animation, accurate period-style screen effects, responsive physical controls, device-specific sound design, and a polished playable game. Games must support pause, restart, exit, scoring or progression, and persistent records using LocalStorage.
This is what $22.50 and 4 hours of Qwen 3.8-Max gets you. 🤯
A full first-person arcade museum, six fictional retro consoles, 41,800 lines of code, and every single station is playable.
Walk up, power it on, and play an original game rendered right on the machine's screen, inside the 3D world.
Prompt below ⬇️
#AI #GenAI #Qwen #CloudComputing #AIInfrastructure