Building with Local AI | DGX Spark x2 Health-Tech | EdTech Need more VRAM. Father of 2. https://xitter.cf/t.co/9MPYYPGGFV 🎧 "Health AI Morning Briefing" 🍎 Podcasts & Spotify
Joined November 2008
- Tweets4.8K
- Following1.3K
- Followers728
- Likes8K
Pinned Tweet
🚨Big model drop!!!
@MiniMax_AI 3.1 Flash now on HF.
Pulling now. Benchmarks soon.
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
AND THERE IT IS. Mic drop. 🎤
3T?! Who is hosting this Titan? @ollama ?
Even an EXL3-1Q would be ~190GB and even agnet prompt-injection wouldn't fit for local-hosting.
Introducing Kimi K3.1 Pro: our most capable 3T-class open-weight model.
🔹 1M-token context, 2.8T total / 104B active params (MoE), Multimodal.
🔹 Full-project coding: games in Unreal, Unity & Godot.
🔹 Excels at parametric 3D modeling in Blender & CAD.
🔹 More reliable at finance work: Accurate modeling with hallucination down by 65%.
🔹 Matching the strongest proprietary models across coding, agentic and knowledge benchmarks.
Kimi K3.1 Pro and K3.1 Pro Highspeed (2–3× faster) are now live on Kimi.ai, Kimi Work, Kimi Code, and the Kimi API.
Open weights by Nov 4, 2026.
🔗 API: platform.kimi.ai
🔗 Tech blog: kimi.ai/blog/kimi-k3-1-pro
Whoa 😲 👀
MiniMax M3.1 Flash is out, and we've open-sourced a DFlash 2 drafter for it under Apache 2.0. It makes generation up to 3.5× faster, and it's lossless. Congrats to @MiniMax_AI on the release!
huggingface.co/incoai/MiniMa…
Wait so we could root a Xaomi or Fire Stick and have this or an Omarchy launcher? @dhh @MichaelGannotti ??
I vibe-coded my TV.
$50 Xiaomi TV Stick now boots into a home screen I built with AI:
→ no ads, no sponsored rows, no tracking
→ my whole apartment in 3D
→ one press and the real lights change
→ “Movie”: lights down, Netflix opens
Both apps are open source. Full setup guide with copy-paste prompts 👇
xitter.cf/0xProbabillity/status/…
Hear me out.
A FOSS firestick OS/launcher - for your TV. Raspberry pi zero or Esp32 or full pi? What’s in a fire stick?
James Aita retweeted
Check out James Aita's nvidia/NVIDIA-Nemotron-3.5-Bolt-74B-A9B-24GB running on 1 DGX Spark with vLLM
Full benchmarks and recipe at: spark-arena.com/benchmark/0b…
James Aita retweeted
Check out James Aita's nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-NVFP4 running on a single DGX Spark with vLLM.
Benchmark and recipe at: spark-arena.com/benchmark/91…
James Aita retweeted
Check out James Aita’s LibertAIDAI/DeepSeek-V4.1-Flash-1017-EXL3-2.9bpw on a 2 nodes NVIDIA DGX Spark cluster with vLLM!
View full benchmark at spark-arena.com/benchmark/c6…
Guys.
We’re gonna get drops from @Kimi_Moonshot , @MiniMax_AI , and @Alibaba_Qwen in the same 24 hours.
Hey @ashxhart does Tensorfold support minimax? If we want some EXL3 serving of Minimax is that possible?
Specialized healthcare model drop. How interesting
Ling-3.0-flash-Sante is now open-source. The 124B healthcare MoE activates only 5.1B parameters per token.
🤖 modelscope.ai/models/inclusi…
🏆 #1 among flash-size models on professional healthcare: Scores 54.29 on HealthBench Professional, up 8.56 points, and ranks #2 overall in the team’s evaluation.
🩺 Better at hard medical reasoning: Reaches 76.92 on AFUMED-ComplexReasoning, 90.61 on AFUMED-Drug, and 94.21 on MedQA-USMLE.
🧩 Multi-agent beats single-agent research: BrowseComp rises from 80.20 to 90.47 when moving from one agent to multi-agent collaboration.
🛠️ A healthcare agent, not just medical QA: Deep research, tool use, and task execution gains support real research and professional workflows.
Get your daily dose of AI news (with a twist of health) wherever you get your podcasts. 6-9 mins - I listen every morning, and it's BANG ON.
podcasts.apple.com/ca/podcas…
With all these crazy new quantization technologies - is anyone trying to get Kimi k3 to fit in a 2-spark setup?
Looks like China has a new AI lab called iflytek- (@iFLYTEKDevs )coming to compete with @deepseek_ai @Alibaba_Qwen and @MiniMax_AI - new model - weights haven’t dropped yet but 293B MoE claims competitive with Deepseek 4.1.
What’s crazy about these guys is they trained ENTIRELY on Chinese silicon. No @NVIDIAAI chips. This is the west’s embargo coming back to bite them - they’ve created a whole new beast that doesn’t run on CUDA.
