Jetson Trial

Hands-on NVIDIA Jetson (Orin/Thor) LLM & edge AI experiments | llama.cpp, real tokens/sec benchmarks | Sharing what actually works

[Jetson Trial @JetsonTrial]

Snapmaker U1 owners — that high-pitched whine from the right panel during printing?

It’s air being pulled through the cavities between the 3rd and 4th filament slots.

I made a passive activated carbon filter that drops the noise and cleans the air a bit. The 3d file is open source (MIT).

Don’t lower fan power though — overheating risk.

Anyone else dealing with this? Other quieting mods you’ve tried?

[Jetson Trial @JetsonTrial]

Seems to be an speed-performance balanced MoE model for Jetson devices

[Liquid AI @liquidai]

Today we release 𝙻𝙵𝙼𝟸.𝟻-𝟸𝟺𝙱-𝙰𝟸𝙱, the largest 𝙻𝙵𝙼𝟸.𝟻 family member. It is an efficient local model for laptops and PCs to work through multi-step tasks inside popular agent harnesses.

> 24B total parameters, 2B active per token
> Context expanded to 131K tokens
> A 128K-token vocabulary
> 33T pretraining tokens
> It is a reasoning model
> Very easy to fine-tune and customize
> Performs competitively with Qwen3.6-35B-A3B, a model with ~50% more total and active parameters, on agentic and tool-use tasks.
> Open base and post-trained weights on @huggingface 🤗

1/

[Jetson Trial @JetsonTrial]

You know you can't be chasing after new models all days.

[Jetson Trial @JetsonTrial]

@Adejare_szn True words

[Jetson Trial @JetsonTrial]

Discouraging...anyway😃

[Jetson Trial @JetsonTrial]

Snapmaker U1 owners — that high-pitched whine from the right panel during printing?

It’s air being pulled through the cavities between the 3rd and 4th filament slots.

I made a passive activated carbon filter that drops the noise and cleans the air a bit. The 3d file is open source (MIT).

Don’t lower fan power though — overheating risk.

Anyone else dealing with this? Other quieting mods you’ve tried?

[Jetson Trial @JetsonTrial]

read the full blog down below😀

https://t.co/iTjFQP4rvx

[Jetson Trial @JetsonTrial]

😂

[Evan Kirstel #B2B #TechFluencer @EvanKirstel]

@Pontifex

[Georgi Gerganov @ggerganov]

Decision models in llama.cpp are now available

The `/v1/systemone` endpoint is available in the latest llama builds. Use it to do Jev-style inference locally, efficiently and privately. Multiple open models are supported with more to come.

https://t.co/D3MUi78kXn

[Jetson Trial @JetsonTrial]

hermes desktop works well on jetson thor. i thought it was not gonna work for aarch64 linux but it does work. looks better than the builtin cli and dashboard interfaces.

[NVIDIA HPC Developer @NVIDIAHPCDev]

Introducing CUDA Rust!

CUDA Rust lets you write GPU kernels natively in Rust, not just launch them from it.

Two paths: cuda-oxide for SIMT kernels compiled to PTX, and cutile-rs for Tile-based programming on stable Rust. Both can catch aliasing errors at compile time.

Technical blog: https://t.co/5mYSRnQkBB

[Jetson Trial @JetsonTrial]

thought "need need"

worked for a while before it indefinitely repeats itself.

dont know if the quantization went wrong or the model. same llamacpp starting arguments worked fine for qwen 35b and orin35b

[Jetson Trial @JetsonTrial]

https://t.co/kvDDcqpzk1

this 35B model bench numbers look pretty- even better than ornith1.5 35B🧐testing

[Jetson Trial @JetsonTrial]

@tvytlx This deserves more traffic than the fake ones.

[Jetson Trial @JetsonTrial]

GM🐦
new day new try

[Jetson Trial @JetsonTrial]

@Elle_Arin Thank you😃

[Jetson Trial @JetsonTrial]

500 followers🎉. Took me about a month.

Thank you — this wouldn’t have been possible without you guys 🙏

How long did your first 500 take?

[Jetson Trial @JetsonTrial]

Once I learn how to drive Hermes Agent, a cheaper/weaker model inside that harness often feels “good enough” compared with paying for the frontier lab model that tops DeepSWE 1.1

[Jetson Trial @JetsonTrial]

Using a coding agent makes my mind feel foggier than pairing with a chat UI and guide the code generation myself.

Cognitive offloading is real. When the agent just does it, I stop holding the problem in my head.

Thinking of going back to the old-school copy-paste loop for a while and seeing if the sharpness comes back.

Do you think ai is degrading your cognitive ability?

[NVIDIA Robotics @NVIDIARobotics]

⚡ Fine-tune and quantize AI models directly on NVIDIA Jetson.

Our new Jetson AI Lab tutorial shows how to use Unsloth and memory-efficient QLoRA to customize models, export quantized GGUF files, and run them locally with llama.cpp.

Follow hands-on examples for:
🔹 NVIDIA Nemotron 3.5 Lightning on Jetson AGX Thor
🔹 Qwen3.5-4B on Jetson Orin Nano

Start optimizing: https://t.co/hDUCiNnQcW

[Duesmoney ⭐️ @Duesmoney]

Movies have a funny way of showing us what we already know but refuse to admit.

Sometimes the character we’re judging is carrying the same flaw we’re trying to hide