TensorFold engine 1.2.1 is out.
A few highlights:
• MiniMax-M3.1-Flash on CUDA across two GB10s, using Mia-AiLab's EXL3 pack.
• Gemma 4 31B on GB10 with Google's MTP assistant: 28.3 tok/s versus 12.2 undrafted, at one stream.
• Qwen3.5/3.6 windows of 1,048,576 tokens on CUDA with YaRN.
• MOSS-Image end to end on Metal, via the CLI.
• Gemma 4 E4B prompt processing 2 to 2.6 times faster on Intel Arc Pro B70.
There are also opt-in GLM long-conversation improvements, server load reporting, Living Weights bulk fact deletion and fixes for lesson failures and Flash Next's long-prompt slowdown on GB10.
Huge thanks to everyone who contributed, tested and reported issues:
@andersjw_ @atasrb @Bizuayeu
@grantoverton @ivanfioravanti @jayleaton
@niklaslenz_ai @2Halves1brain @wesley4546
And on GitHub: AbrahamCete, akol1, BlivionIaG, BobClawblaw, gilby, Jenpo, jschmied, kieranscotney, laudney, Mirrdhyn, ppardus, russellmm and ThomasHalwax.
Upstream fix credits: florianbrede-ayet and mikolaj92.
Thank you for the time, code, testing and hardware access you keep sharing with TensorFold. Apologies if I've missed anyone.
@petruspennanen @plotarmordev @MiaAI_lab @volatilemarkts
tensorfold.dev
Anders retweeted
Do you know the meaning of the @NousResearch logo?
The girl is called the "Nous Girl". She was drawn in July 2023 by Nous's first artist, who goes by John Galt. Yes, the Atlas Shrugged rebel
She stands for the rebellion of decentralized AI, and for keeping our humanity in the face of censorship and machine intelligence
The words around her: "Rebellion to tyrants is obedience to God."
That line goes back to 1776. Jefferson, Franklin and Adams put it on their proposed Great Seal of the United States. Congress said no, so Jefferson put it on his own personal seal instead
Nous pointed it at AI. Today's tyrants are closed labs and gatekeepers deciding who gets access to intelligence. The rebellion is open weights, open agents and open infrastructure.
That's why I build on Hermes
Wearing it proudly 🏴
Now that TensorFold has joined NVIDIA Inception, we now have early access to the next Nemotron. 🚀
We’re using this access to build and test ahead of release, bringing 0day support to TensorFold on both MLX and CUDA.
@NVIDIAAI @abheer @volatilemarkts @MiaAI_lab
Anders retweeted
GLM 5.3 Flash on 2x DGX Sparks v1.10 is out ⚡️
Lots of PRs merged and bugs squashed!
Here are some highlights:
- Works with Claude Code & Anthropic SDKs directly
- Much faster at very long chats (500k+ tokens)
- Opt-in disk cache: long chats resume in 0.2 s instead of 11 s, even after a restart
- Sampling is now as fast as greedy decoding
Full changelog here:
mia-ai.net/models/GLM-5.3-Fl…
Anders retweeted
As reported in the @WSJ, we have raised a Series B to bring Hermes Agent to new frontiers (and, yes, build a mobile app)
Thank you to our investors including @nvidia @M12vc @SamsungNext @robotventures @usv @ycombinator @MenloVentures, among others
wsj.com/pro/venture-capital/…
sharing my thoughts on our recent fundraise.
we did alright for a couple of goofballs
nousresearch.com/a-note-on-o…
Anders retweeted
What a generational fumble from Mistral.
The French startup was meant to be the sovereign AI alternative in Europe.
Instead, it moved away from building models to applications and infrastructure.
It decided to move away just at the time when open source models became good enough and cheap enough to be viable.
Now Aleph Alpha has swooped in with a cheap, high-performing, open-source model.
The German company has raised about 15% of the cash that Mistral has riased.
Anders retweeted
Replying to @ashray_malhotra
Someone told me that Hermes thing everyone talks about lets you control what your agent has access to and how it uses it. Idk sounds kinda crazy but I might check it out
Anders retweeted
Can confirm, this absolutely rips
This is THE update everyone was waiting for ⚡️
MASSIVE speed gains for GLM 5.3 Flash EXL3 running on 2x DGX Sparks
- 29% faster on average for prose (!)
- 20% faster on average for code
Prose decode speed is now around 36 tok/s!
This is the FASTEST GLM 5.3 Flash setup yet for two dgx sparks, and is now my go-to.
Expect further improvements!
Get it here:
github.com/MiaAI-Lab/GLM-5.3…
Anders retweeted
I probably shouldn’t be showing this yet…
But Hermes is about to get a command center in your pocket.
Every gateway.
Every machine.
One fleet.
Hermes Fleet for iOS. 🪽
This is gonna get ridiculous.
Anders retweeted
The new Qwen3.8 2.4T A95B already on @NVIDIAAI DGX Spark - slow but running. Impressive work by @WescheNex1q
Qwen3.8 2.4T-A95B recipe
4× DGX Spark
UD-Q1_0
64K context
RoCEv2 RDMA
Native MTP3
7.76 tok/s
The exact reproducible recipe, pinned runtime, hashes, launch scripts, and evidence are now public:
github.com/Weschera/Qwen3.8-…
Anders retweeted
There is a Deepseek v4 Flash version for everyone of you! 😃
Delayed but here it is. DeepSeek-V4-Flash-0731 in GGUF at 2.88 bits per weight, 95 GiB, fits and runs on one 128 GB box. Config-I hybrid quant: 2-bit experts, WHT-rotated 3-bit attention, routing kept at full precision. Wikitext PPL 15.6.
Model: huggingface.co/thetom-ai/Dee…
Operating guide, including the backend traps that will bite you: github.com/TheTom/offlabel/b…
Thanks Starlink for taking 3 days to upload!
So much in the Herald Release of Hermes Agent
- Voice Activated Chats (Replace your google home and alexas!)
- Plugins and Kanban for the Desktop GUI App
- Agent2Agent Protocol
- Outbound Webhooks
- Massive performance optimizations and token efficiency milestones reached
- Buzz Integration
- Nous Portal's 20% Discount on all models
- Way more reliable local and open models
- New office/productivity skills
- Grounded research skill
- Secrets lockdown with Ironproxy
and like 500 other things. Check them all out!
Anders retweeted
Congratulations @deepseek_ai on the launch of DeepSeek V4 Flash 0731!
Get started today by running the model locally via the @UnslothAI GGUF. 👇
DeepSeek V4 Flash 0731 can now be run locally! 🐳
Run DeepSeek V4 Flash lossless 4-bit on 168GB RAM and 3-bit on 110GB RAM.
V4 Flash 0731 outperforms V4 Pro. Run via Unsloth or llama.cpp. Smaller quants coming today.
Guide: unsloth.ai/docs/models/deeps…
GGUF: huggingface.co/unsloth/DeepS…
Anders retweeted
We’ve decided to open-source a multi-agent harness we use internally at YC.
We call it “QM” and it’s meant to be easy to customize, like Hermes or OpenClaw, but useful for a whole company. We use it across accounting, legal, events, and engineering (including building QM itself!).
The whole project is under an MIT license. It is cloud-first and has Slack and web UI natively.
At only 118b param, Laguna S 2.1 achieved 78.5% on SWE-bench Multilingual, beating models up to 200x its size.
- Open weight
- 1M context
- Runs on a single DGX Spark
Available now through OpenRouter. Welcome @poolsideai Laguna to Pi!
Today we're releasing Laguna S 2.1, our most capable model to date.
It's a 118B total parameter Mixture-of-Experts model with 8B activated per token, a context window of up to 1M tokens, and thinking and no-thinking modes.
Capable enough to hold its own against models many times its size. Small enough to run on a single @NVIDIAAI DGX Spark.
Laguna S 2.1 is fully open under OpenMDW-1.1, with weights available today on @huggingface
poolside.ai/blog/introducing…







































