Ash Hart
@ashxhart
Doing my bit to bring local AI to everyone | building | | | Thoughts are my own.
474 Following 8.8K Followers
You can use TensorFold without opening a terminal.
The Mac app has the engine built in. Choose Full mode, load a compatible model and open a chat.
I've put the setup steps here:
Show more
@TechMDAI @MiaAI_lab We are just doing our bit to push local AI forward dude!
TensorFold 1.0.12 is out with Nemotron 3.5 Super VL 120B vision on 1 or 2 DGX Sparks, Qwen3.8-27B vision and Nemotron oQ on Macs, LFM2.5-24B, plus source previews for GLM-5.3-Flash NVFP4 across Sparks and Intel Arc.
Thanks to the 20 people in 1.0.12:
alenperic, BHCC2025, Bizuayeu, BlivionIaG, BobClawblaw, cwschroeder, harrisonfriia, heitke, jayleaton, jishnuvenugopal, jschmied, kelchm, liumorrisclaw, MiaAI-Lab, Mirrdhyn, mrv777, philip-pentatonic, plotarmordev, sethforprivacy, ThinkOffApp.
And thanks to jkuepker and SvangenStudios for runs on their own hardware, and grantoverton, whose oQ work this release builds on.
@petruspennanen @plotarmordev @volatilemarkts @MiaAI_lab
Show more
I asked @Bot to fire itself from my machine and it built me a private AI using TensorFold then I switched off Wi-Fi.
The AI kept running. 🤯
No cloud. No per-token API bill. Just my laptop.
Prompt in the comments 👇🏼
🚀
Show more
Did somebody say TensorFold for Apple Watch? 👀
A little pov of what’s coming along with an iOS and iPadOS app 😃
@petruspennanen @volatilemarkts @MiaAI_lab @plotarmordev
Show more
Nvidia Nemotron 3.5 Super VL is live with 0Day Support on TensorFold. 💪🏻
@MiaAI_lab @NVIDIAAI @volatilemarkts @petruspennanen
TensorFold now has a "Beta" macOS app to launch the engine, so you don't need to jump into the CLI.
It's also under /app in the git repo if you want to drop any PR's against it.
Show more
This is excellent!
Introducing GLM-Legal
Built on GLM-5.3-Flash, GLM-Legal matches frontier models on mainstream legal benchmarks.
Designed for real legal workflows: contract review, evidence analysis, compliance, and more.
Tech Blog:
Apply for API Access:
Try AutoClaw:
Show more
Living Weights is helping to change the game of local AI, I hope.
So we've been testing Living Weights with you with lessons like "I like blue. My eyes are brown." then "My eyes are green." And last "My eyes were brown until 11 October 2026; now they are green."
Then removing one of the lessons and asking for the color. This has already helped us but it's a bit ad hoc.
So I just started cooking a generalised Living Weights logical consistency checker. It has objects with attributes, values and time and relations between them.
A genetator then createst tests of desired size and complexity, plus the Lean code to check the results automatically.
And if they get too logical for us we can always add some temp in the model 😄
Show more
Phenomenal 👌🏻👏🏼
customer Installation day at @seesa_ai 😍
We are installing a fleet of DGX Sparks on prem for a FTSE100 company that has billions of pounds of assets under management and cannot risk their data in the cloud.
With Seesa’s custom local models, harness, memory engine and enterprise security they don’t have to choose anymore!
Enterprise, Local Ai. Nice.
Show more
Living Weights is getting more love.
What’s everyone cooking today 💪🏻
Intel B70 owners we got you 🙌🏻
On bosse sur le portage XPU et le support de la carte Arc Pro B70 sur #TensorFold# version Zig, avec pas mal d'optimisations disponibles sur la table, notamment au niveau XMX et OpenCL. Objectif : 2400 tok/s en prefill et 100+ en decode @ashxhart @fragtalife @IntelFrance #LocalAI#
Show more
Let’s go 🚀
Messing with Tensor Fold today and so far got it running on NVFP4. What quant for EXL3 are you seeing success with right now ?
Deepseek V4.1 flying at 130tks decode 😦
After another 100 commits, recipe/configs ready for 4x sparks:
Deepseek V4.1 Tensorfold Zig rebase
~130 tok/s decode
~5k prefill
Thanks goes to @ashxhart and @jayleaton for their inspirational work!
Show more
Lots of you asked for a TensorFold Discord, so it's finally here.
Come share model recipes and benchmarks, report bugs, get setup help or join the contributor discussions.
Come say hello:
@MiaAI_lab @volatilemarkts @petruspennanen @plotarmordev
Show more
The full GLM 5.3 EXL3 hitting ~95tp/s decode 🤯🤯🤯🤯🤯
Optimized TensorFold for GLM5.3 EXL3 and achieved 87-95 tok/s on 4x RTX PRO 6000! Speculative decoding is MTP.
Uncensored SOTA models at home are starting to become a reality🔥
Show more
TensorFold on the other side of the planet!
Local AI is going to take off 🚀
Optimized TensorFold for GLM5.3 EXL3 and achieved 87-95 tok/s on 4x RTX PRO 6000! Speculative decoding is MTP.
Uncensored SOTA models at home are starting to become a reality🔥
Show more
