one cloud for compute, inference, and agents Question: discord.gg/d6RhER7ySx Japan: @gmicloudJapan
Mountain View, CA
- Tweets 3,637
- Following 62
- Followers 10,183
- Likes 3,402
Pinned Tweet
40% off DeepSeek V4.1 Flash 1017
10 days, unlimited, run thru GMI API 👇
(holiday season is coming, what model you want discounts for? )
11
4
46
12,543
we asked kimi k3.1 pro @Kimi_Moonshot to create an aquarium of creatures from the cthulhu mythos
ran it for 4h30m and cost about $39
we wanted to push it further, but it had a hard time realizing the details
4
1
26
805
GMI Cloud is now listed on the a16z speedrun marketplace!
a16z speedrun companies can unlock up to $5,000 in inference credits, plus 20% off dedicated endpoints and 300+ models behind one API.
Available now to the first 25 qualifying a16z speedrun companies 👇
2
2
10
442
The other half of Day 4: templates and a service.
gmi sandbox templates builds a template from any container image. gmi serve is a small HTTP service. It answers 401 without its token and a real chat through GMI with it.
Short waits are shortened.
Day 4 of @zkasuran x GMI: AgentBox sandboxes from the terminal.
gmi sandbox run starts a GMI sandbox, runs your command inside it and cleans up. Exec passes the exit code through. Files go in and out.
Recorded live. No waits are shortened.
1
7
347
Day 4 of @zkasuran x GMI: AgentBox sandboxes from the terminal.
gmi sandbox run starts a GMI sandbox, runs your command inside it and cleans up. Exec passes the exit code through. Files go in and out.
Recorded live. No waits are shortened.
1
4
7
651
Congrats to MiniMax on the open-weight release of MiniMax M3.1 Flash!
We had it live on GMI Cloud from day one. It's now also available on @OpenRouter, served by GMI Cloud.
Use it through the GMI API or pick GMI as the provider on OpenRouter 👇
MiniMax M3.1 Flash weights are now public. 428B parameters total, 23B activated per token.
Agentic performance: A generational leap in agentic tasks, coding, and cybersecurity — 76.2 on BankerToolBench, 73.7% on DeepSWE v1.1, 61.7 on OSWorld V2, 29.6 on Agents' Last Exam, and 82.2 on CyberGym.
Efficient by design: MiniMax Sparse Attention and DSpark speculative decoding improve inference efficiency.
Hugging Face:
huggingface.co/MiniMaxAI/Min…
ModelScope:
modelscope.cn/models/MiniMax…
API: platform.minimax.io
M Plan: minimax.io/m-planMiniMax
MCode for desktop: agent.minimax.io/download
6
4
31
924
We're proud to support @UtopaiStudios.
Utopai X, the studio's video model, ranks #2 on the Artificial Analysis Text-to-Video Leaderboard With Audio.
Trained and served on GMI Cloud.
How they built it on GMI 👇
Meet the all-new PAI & Utopai X, highest-ranked model from a U.S.-based company on Artificial Analysis’ Text-to-Video Leaderboard with Audio. Utopai X works inside PAI, our production intelligence engine, where workflow and creative decisions connect across pro-level productions.
4
3
22
864
What the live check found.
gmi probed the 17 read operations that need no ID: 15 answered ok and 2 IAM reads need a different key type. The other 76 were not run because they write or need an ID.
Sandbox and Cluster need a compute key. The inference key is rejected.
Day 3 of @zkasuran x GMI:
every GMI API operation as a command.
93 operations from GMI's four published specs, generated rather than hand written. gmi api list shows them, --dry-run previews a request and gmi api check probes the safe reads.
Recorded live. One long wait is shortened.
1
2
9
532
GMI Cloud retweeted
Tomorrow. 🎃
6 weeks of building. 1 haunted demo day. SCALE at GMI
Cohort 2 takes the stage at AGI House, Hillsborough.
🗓️ Fri, Oct 23 · 12–8 PM PT
@gmi_cloud, @nvidia
2
3
13
633
Day 3 of @zkasuran x GMI:
every GMI API operation as a command.
93 operations from GMI's four published specs, generated rather than hand written. gmi api list shows them, --dry-run previews a request and gmi api check probes the safe reads.
Recorded live. One long wait is shortened.
1
3
20
1,254
amazed by how far open-weight models have come in just 5 months
we used the same prompt from our DeepSeek comparison back in May
went from beginners at gaming and 3D world simulation to pros
5
3
33
1,579
our post in May:
DeepSeek V4 Pro vs Xiaomi MIMO V2.5 Pro
MIMO V2.5-Pro wins by:
1. ~1 min faster than DeepSeek-V4-Pro
2. cleaner aesthetic
Our reasoning: MIMO V2.5 Pro is natively multimodal, so it has a stronger sense of what good visuals look like.
Curious to see how DeepSeek V4 Vision will change that
2
216
We are proud to support the @NVIDIAAI OpenShell Reference Architecture and bring it to our infrastructure.
Building on GMI Cloud’s investment in the NVIDIA Vera Rubin platform: model compute, governed agent execution and hardware-level security are all on one cloud.
Agents can work with real data, tools and services, with clear permissions, runtime controls and a full record of what they do.
AI agents need fast tool calls and security controls they can’t rewrite.
NVIDIA OpenShell Reference Architecture combines the compute to run agents at scale with runtime controls that govern their access to tools and data.
See how the hardware and software work together: nvda.ws/4BiJBfY
Made with AI
3
4
19
21,763
read more: gmicloud.ai/en/blog/gmi-clou…
149