#LLAMAcpp
LightNews LightNews
plushie.holdings
im gonna do something really funny if u give me two hours to port this moe architecture to llamacpp
sungkim.bsky.social Sung Kim @sungkim.bsky.social · 12d
Yandex released the weights of a model (AliceAI-T5-35B-A0.6B-Base) used to generate AI answers in Search.

Blog: medium.com/yandex/meet-...
Model: huggingface.co/yandex/Alice...
September 14, 2026 at 8:36 PM
deadprogram.com
First up @goinggo.net lightning talk about kronkai.com here at @gophercon.com
#gophercon #golang #llm #llamacpp
August 6, 2026 at 8:36 PM
jamiebeach.bsky.social
llamacpp settings I used : gist.github.com/jamiebeach/3...
llamacpp launcher powershell script for Ternary Bonsai 2 model
llamacpp launcher powershell script for Ternary Bonsai 2 model - start-bonsai.ps1
gist.github.com
September 23, 2026 at 9:00 PM
yzma.ai
yzma 1.28 is out! Run local AI models from Go with no CGo. Support for the new @hf.co llama.cpp v0.5.0, OpenVINO, autoselect CUDA, updated benchmarks, more WASM support.

Go get it right now!

yzma.ai
#golang #llamacpp #sovereignAI #localAI
yzma
yzma lets you write Go applications that use llama.cpp for local inference with full hardware acceleration.
yzma.ai
September 24, 2026 at 1:29 PM
simon.fedi.simonwillison.net.ap.brid.gy
DeepSeek R1 appears to be a VERY strong model for coding - examples for both C and Python here: https://simonwillison.net/2025/Jan/27/llamacpp-pr/
ggml : x2 speed for WASM by optimizing SIMD
PR by Xuan-Son Nguyen for `llama.cpp`: > This PR provides a big jump in speed for WASM by leveraging SIMD instructions for `qX_K_q8_K` and `qX_0_q8_0` dot product functions. > > …
simonwillison.net
January 27, 2025 at 6:35 PM
deadprogram.com
Less than 2 weeks until Embedded World & I will be at large on the expo floor! Let's chat about all things Go with microcontrollers, computer vision, & machine learning. Just look for Gopherbot.

#golang #tinygo #ew26 #embedded #computerVision #ml #openCV #llamacpp #yzma
Gopherbot blinking
February 26, 2026 at 10:46 AM
kenthropic.com
While the Svelte community was buzzing about Apple's App Store leak, here's the real gem: llama.cpp's new official WebUI – built with Svelte/SvelteKit! Run any of 150k+ GGUF models with a gorgeous interface. Fully local, fully open source 🚀 #Svelte #SvelteKit #LlamaCpp 👇
github.com/ggml-org/lla...
llama.cpp WebUI interface showing chat window with Qwen3 VL 30B model loaded, displaying clean minimal design with text input, file attachments, and voice input support
November 5, 2025 at 5:07 PM
iballantyne.bsky.social
How fast is #gemma3 270M out of the box on a @raspberrypi.com 5? About 30 tok/s on CPU for the Q4_0 model using #ollama. Tried using #llamacpp, got about 32 tok/s. 🤯

Extremely promising for edge compute with no accelerator! Model is only ~250Mb on disk. Can't wait to fine-tune for my IoT projects 👨‍💻
August 15, 2025 at 12:39 AM
ashanti.pds.witchcraft.systems
update: this is fucking cursed, how did meituan ever think releasing this architecture was a good idea

the sparse attention is completely fucking busted for more than 32 keys for all the llamacpp forks i've tested, trying to monkeypatch the attention to actually work now
ashanti.pds.witchcraft.systems Ash @ashanti.pds.witchcraft.systems · 2d
sometimes i feel like the only person in the world interested in working on longcat flash lite sparse, it's just such a perfect size, and has the balls to be middling in performance _and_ an explicitly non-reasoning model while also having prefill optimisations over DSA to make it cheaper to run
September 25, 2026 at 11:32 PM
personhood.removal.surgery
ROCm fucks, but the downside is IME that when it randomly eats shit it knocks the entire GPU off the bus instead of just killing llamacpp ;_;
May 8, 2026 at 7:45 AM
adriend.linuxtricks.fr
Je me suis encore amusé cet après-midi (mauvais temps oblige).
Un LLM local ce'st bien mais sans recherche web c'est pas top.
On s'était amusé en live avec des recherches Wikipedia.
Là c'est tout le Web qui est interrogé :)
#python #streamlit #duckduckgo #llamacpp #qwen35
March 28, 2026 at 5:39 PM
oof.dere.systems
i managed to make llamacpp have somewhat nice dx while staying pretty low-level
May 14, 2026 at 8:50 AM
frankel.ch
#OpenCode with #Llamacpp — How it Works in Practice

https://medium.com/rigel...
July 11, 2026 at 5:00 PM
hydroxide.dev
I've worked with Martin in Singapore before, and just discovered that we both share the same (ancient) GTX 1080 setup. He has a blog post detailing how he managed to get 30B MoE model up and running via llamacpp:

mdda.net/blog/tech/dl...
Running ~30B MoE models on an old GTX 1080
( Fedora 42, llama.cpp )
mdda.net
June 22, 2026 at 5:51 AM
webbrain-one.bsky.social
We take Security seriously at WebBrain #llm #ai #offlineAI #qwen #gemma #gemma4 #llamacpp #openllama #opensource #chrome #ollama #
June 27, 2026 at 1:01 PM
stephanjanssen.be
Experimental #LlamaCPP Code Completion Project for @jetbrains.com IDE's using any GGUF model 🤩 🔥
GitHub @ github.com/stephanj/Lla...
January 31, 2025 at 3:34 PM
yzma.ai
Very cool, our latest release is in the new Golang Weekly newsletter!

🤖 yzma 1.26: Local LLM Inference from Go, Now in the Browser

golangweekly.com/issues/617

#golang #llamacpp #tinygo #wasm #webgpu
Building cgo code without a C compiler?
Go Weekly — Issue #617
golangweekly.com
September 11, 2026 at 4:05 PM
stephanjanssen.be
Qwen 3 @ chat.qwen.ai or API access via HuggingFace and LlamaCPP or ModelScope... 🔥👀
April 29, 2025 at 6:18 AM
hybridgroup.com
yzma 1.4 is out, the last release of the year. Support for split models and a few new features that just got added to llamacpp too. Enjoy high performance local inference from Go!

github.com/hybridgroup/...

#golang #llamacpp
GitHub - hybridgroup/yzma: Write Go applications that directly integrate llama.cpp for local inference using hardware acceleration.
Write Go applications that directly integrate llama.cpp for local inference using hardware acceleration. - hybridgroup/yzma
github.com
December 29, 2025 at 6:46 PM
a.rapeable.dog
i need to try out one of those llamacpp forks with turboquant i guess
April 16, 2026 at 7:00 PM
ashanti.pds.witchcraft.systems
I really should try the model again but in at least 4 bit this time (with the embedding table in at least q8), but I don't know if it'll do too well, I'm more interested in ling-3.0-flash-vl getting better llamacpp support soon because that one looks really good too
September 22, 2026 at 9:57 PM
yzma.ai
"Building Gemma 4 Local-Powered LLM Apps with Go and Yzma" new post from Vladimir Vivien

medium.com/@vladimirviv...

#golang #llama #llamacpp #gemma #yzma
Building Gemma 4 Local-Powered LLM Apps with Go and Yzma
Using Go to build an local LLM-powered weather application with on-device inference and tool calling using Gemma 4
medium.com
April 15, 2026 at 7:35 AM
hailey.at
models like gemma or qwen are probably the best bet since tool calling is so important. they also have very good reasoning, and once llamacpp lands mtp support the tps should be quite good
May 6, 2026 at 5:30 PM
yzma.ai
Just released yzma 1.12 with a critical fix, and some additional test coverage.

go get it!

github.com/hybridgroup/...

#golang #llama #llamacpp #yzma #kronk
GitHub - hybridgroup/yzma: Go with your own intelligence - Go applications that directly integrate llama.cpp for local inference using hardware acceleration.
Go with your own intelligence - Go applications that directly integrate llama.cpp for local inference using hardware acceleration. - hybridgroup/yzma
github.com
April 15, 2026 at 9:33 PM
ripter.bsky.social
I built a CLI to make it easier to run local LLMs with llama.cpp.

A few weeks later I have another tool using it to read my email and apply labels with a local model.

No credits, just scripts and a model on my laptop.

This is what I wanted local LLMs for.

#llm #localllm #llamacpp #devtools
March 14, 2026 at 12:29 AM