#llamafile — Public Fediverse posts

Finally (maybe) a usable local LLM that works on normal (and old) CPUs, not like a gimmick. This one on a 9th gen Intel, 16GB RAM with built-in GPUs (and with zram active). This specific Gemma model seems to be doing quite well in terms of speed, is not too dumb, and it apparently uses the Vulkan API out of the box. It also doesn't seem to spike the CPU at 100%, although it has crashed when the context was long. Web UI has improved. Still needs lots of testing though...

#ai #llm #llamafile

#Mozilla #AI Releases #Llamafile 0.10.4, their solution for easy-to-use #LLM as a single file that work across hardware and operating systems. With Llamafile 0.10.4 is now #Transcribefile, as a new piece built off their recently announced Transcribe.cpp project
Llamafile 0.10.4 also updates against its Llama.cpp upstream build, brings a few improvements to its #Vulkan #API and #AMD #ROCm acceleration handling, HTTPS download support, and pledge/SECCOMP sandboxing support
https://www.phoronix.com/news/Llamafile-0.10.4

What if the soon to be released @silex Desktop shipped a way to fine-tune your own small, open AI locally?

A ~1B open model could be fine-tuned right inside Silex to learn how the user uses Silex and imitate via MCP. Runs on modest hardware, your data never leaves your machine

The stack: #Luciole, the model by #OpenLLMFrance / @LINAGORA, #llamafile by @mozilla

I'd be happy to find Silex users who want to co-build this, I have no idea how to make it happen 😅

In? 👇

#foss #LocalAI #llamafile

(more Linux and FOSS news in previous posts of thread)

Mozilla AI Releases Llamafile 0.10.4 With New Transcribefile Built On Transcribe.cpp:
https://www.phoronix.com/news/Llamafile-0.10.4

Home Assistant Matter Server 9.1.0 adds support for Matter 1.6.0:
https://alternativeto.net/news/2026/7/home-assistant-matter-server-9-1-0-adds-support-for-matter-1-6-0/

Zed Editor 1.11.3 Brings Dedicated Git Views, Sonnet 5 Support, and 60+ Bug Fixes:
https://www.linuxcompatible.org/story/zed-editor-1113-brings-dedicated-git-views-sonnet-5-support-and-bug-fixes/

Design tool Penpot brings background blur, better WebGL rendering & design token upgrades:
https://alternativeto.net/news/2026/7/design-tool-penpot-brings-background-blur-better-webgl-rendering-and-design-token-upgrades/

GitLab 19.2 brings Duo CLI, custom AI flows, and scheduled policies GA:
https://alternativeto.net/news/2026/7/gitlab-19-2-brings-duo-cli-custom-ai-flows-and-scheduled-policies-ga/

Gitea 1.27.0 patches major security flaws and enhances Actions:
https://alternativeto.net/news/2026/7/gitea-1-27-0-patches-major-security-flaws-and-enhances-actions/

Forgejo 16.0 adds SSRF protection, granular notifications, and refined review tools:
https://alternativeto.net/news/2026/7/forgejo-16-0-adds-ssrf-protection-granular-notifications-and-refined-review-tools/

PostgreSQL 19 Beta 2 Released: Native Graph Queries, Unified REPACK, and Feature Freeze:
https://www.linuxcompatible.org/story/postgresql-19-beta-2-released-native-graph-queries-unified-repack-and-feature-freeze/

Rust 1.97.1 Drops Exactly One Week After 1.97.0 to Fix Critical Miscompilation:
https://www.linuxcompatible.org/story/rust-1971-release-critical-llvm-miscompilation-fix-lands-one-week-after-1970/

Python 3.15 Beta 4 Ships With Built-in frozendict, Lazy Imports, and UTF-8 Defaults:
https://www.linuxcompatible.org/story/python-315-beta-4-released-lazy-imports-frozendict-and-utf8-defaults/

OpenClaw v2026.7.1 released with Control UI overhaul, easier setup, new GPT, Codex, Tencent, Claude, etc. models available:
https://docs2.openclaw.ai/releases/2026.7.1

FosseryWeb and page-builders update:
I added article page generation support to page-builders, and ordered list, strikethrough text conversion support to its converter component. Also regenerated the existing article pages, to fix inconsistencies in the code, plus included the title and date of the article in the downloadable ODT and Markdown versions.
https://fosseryweb.codeberg.page/articles/
https://fosseryweb-min.codeberg.page/articles/
https://codeberg.org/fosseryweb/page-builders

(more FOSS news in comment)

#WeeklyNews #OpenSource #FOSSNews #OpenSourceNews #FOSS #News #Llamafile #AI #HomeAssistant #Zed #Penpot #GitLab #Gitea #Forgejo #PostgreSQL #Rust #Python #OpenClaw #FosseryWeb #FosseryTech

I bet that in 2030 we'll be using claude or gpt only to setup local models with llamafile, optimized for specific tasks on a specific laptop or phone

> The only thing Internet Explorer is good for is downloading Firefox

@mozilla #llamafile is a #foss alternative to VC funded #ollama

#ai #llm #slm #fantasticfour #enshitification #OpenSource

Neste vídeo, mostro como criar uma IA local portátil no Linux utilizando o Llamafile e um modelo no formato GGUF, armazenados em um pendrive ou SSD externo.

Link: https://youtu.be/UELV_vluOPQ

#llamafile #ibmgranite #ialocal #tech #ferramentaia #iagratis #InovacaoTech #IAPortatil #linuxmint #linux

Hmmm... 🤨

#Llamafile: Run #AI Models Locally on Your PC with Just One File https://firethering.com/llamafile-run-ai-models-locally-one-file/ #LLM #GenAI

Encántame este vídeo de Mancomún sobre Llamafile! Aprende a executar modelos de IA de xeito sinxelo no teu propio ordenador, con explicacións claras e ligazón ao artigo. Perfecto para quen quere probar IA local e software libre. Anímate a experimentar! #Llamafile #IA #InteligenciaArtificial #SoftwareLibre #Mancomun #Galego #PeerTube
https://v.eurorede.com/videos/watch/22adee43-95aa-4eb5-aa8d-4691a2fa6193

(Linux news in previous posts)

FOSS NEWS

Mozilla is packing new features into Firefox:
https://betanews.com/article/mozilla-is-packing-new-features-into-firefox/

Is Firefox getting a new logo? Mozilla’s socials suggest so…:
https://www.omgubuntu.co.uk/2026/03/is-firefox-about-to-get-a-new-logo

Signal introduces group member labels, allowing you to specify your role or responsibility:
https://alternativeto.net/news/2026/3/signal-introduces-group-member-labels-allowing-you-to-specify-your-role-or-responsibility/

Blender 5.1 Open-Source 3D Graphics Software Released with Many New Features:
https://9to5linux.com/blender-5-1-open-source-3d-graphics-software-released-with-many-new-features

LibreOffice 26.8 To Add A Donation Banner To Its Start Center:
https://www.phoronix.com/news/LibreOffice-26-Donation-Banner

Mozilla Releases Llamafile 0.10 To Enhance Their AI Offering For Easy-To-Use LLMs:
https://www.phoronix.com/news/Mozilla-AI-Llamafile-0.10

Ollama integration brings all models to OpenClaw platform:
https://alternativeto.net/news/2026/3/ollama-integration-brings-all-models-to-openclaw-platform/

OpenShot 3.5 is (yet again) the ‘biggest’ and ‘fastest’ release ever:
https://www.omgubuntu.co.uk/2026/03/openshot-3-5-released

Bitwarden Send improves security with email verification for paid users:
https://alternativeto.net/news/2026/3/bitwarden-send-improves-security-with-email-verification-for-paid-users/

SuperTux 0.7.0 Arcade Game Is Out with Complete Level Design, Revamped Graphics:
https://9to5linux.com/supertux-0-7-0-arcade-game-is-out-with-complete-level-design-revamped-graphics

PlayStation 3 emulator RPCS3 gets easier to use with Steam:
https://www.gamingonlinux.com/2026/03/playstation-3-emulator-rpcs3-gets-easier-to-use-with-steam/

OpenTTD devs clarify store changes with Transport Tycoon Deluxe re-release as Atari contribute server funding:
https://www.gamingonlinux.com/2026/03/openttd-devs-clarify-store-changes-with-transport-tycoon-deluxe-re-release-as-atari-contribute-server-funding/

GNUnet 0.27 Released For Those With "Some Reasonable Pain Tolerance":
https://www.phoronix.com/news/GNUnet-0.27-Released

Immich 2.6 improves map side panel, asset viewer, shared link slugs & presets, and more:
https://alternativeto.net/news/2026/3/immich-2-6-improves-map-side-panel-asset-viewer-shared-link-slugs-and-presets-and-more/

(more FOSS news in comments)

#WeeklyNews #OpenSource #FOSSNews #FOSS #OpenSourceNews #News #Firefox #Signal #Blender #LibreOffice #Llamafile #Ollama #OpenShot #Bitwarden #BitwardenSend #SuperTux #RPCS3 #OpenTTD #GNUnet #Immich #Browser #WebBrowser #VideoEditor #VideoEditing #ContentCreation #Gaming #FOSSGame #FOSSGaming #OpenSourceGame #PlayStation #PlayStation3 #FosseryTech

#Mozilla Releases #Llamafile 0.10 To Enhance Their #AI Offering For Easy-To-Use LLMs
Llamafile is a https://Mozilla.ai project to distribute and run large language models as a single file. With a single Llamafile you can run the #LLM across platforms and with varying hardware support. Their intentions are on making LLMs more accessible and convenient to both developers and end-users.
https://www.phoronix.com/news/Mozilla-AI-Llamafile-0.10

Mozilla Releases Llamafile 0.10 To Enhance Their AI Offering For Easy-To-Use LLMs - Phoronix

｢ Llamafile is a https://Mozilla.ai project to distribute and run large language models as a single file. With a single Llamafile you can run the LLM across platforms and with varying hardware support. Their intentions are on making LLMs more accessible and convenient to both developers and end-users ｣

https://www.phoronix.com/news/Mozilla-AI-Llamafile-0.10

#ai #llamafile #llm #mozilla

Mozilla rilancia Llamafile con la versione 0.10: supporto immagini, GPU Metal e CUDA, Whisper e Stable Diffusion integrati. Un passo avanti per rendere i modelli linguistici più accessibili. #Mozilla #Llamafile #Linux #OpenSource

https://www.linuxeasy.org/mozilla-rilascia-llamafile-0-10/?utm_source=mastodon&utm_medium=jetpack_social

llamafile: Distribute and Run LLMs with a Single File

https://github.com/mozilla-ai/llamafile

#HackerNews #llamafile #LLMs #MozillaAI #AItools #MachineLearning #DistributeRun

🦙 llamafile returns!

Mozilla.ai is adopting llamafile to push forward open, local, privacy-first AI.

We’re refreshing the codebase and rebuilding the roadmap with feedback from the community.

Share your ideas on GitHub, Discord, or Hacker News—and help shape the next phase of llamafile.

🔗 https://blog.mozilla.ai/llamafile-returns/

#MozillaAI #llamafile #opensource #LocalAI #AIcommunity

At DjangoCon US 2025 in Chicago, more than one person shared the workflow of dictating their articles or slide notes to a template using mobile apps 🎙️

I was experimenting with https://huggingface.co/Mozilla/whisperfile, and it seems to work well on my PC 🔴

It occurred to me that it could be used to add live captioning to meetups or small conferences that can't afford live captioners as good as the one we had at DjangoCon 💡

Have any of you done any experiments?

#Captioning #Conference #Whisper #llamafile

@naturzukunft

There aren't that many choices to manage exploding information complexity so something must happen.

Even if it is not the semantic web as we knew it, it must solve similar problems (and hence look similar - as form follows function).

Removing frictions is key imho. See e.g. how using complex LLM models is now down to just one #llamafile

Semantic web tooling (at least the public / open source part) never reached a sweetspot for broad adoption, always somewhat aloof

@rzeta0

In #BYOTA, we use #llamafile to calculate *sentence embeddings*.
If you don’t know what embeddings are, just think about them as numerical descriptors of your Mastodon statuses, which are closer (as in two cities’ coordinates being close on a map) the more semantically similar their respective statuses are. We’ll get back later to this with a more visual description. If you are interested in embeddings and wanna delve deeper, see https://vickiboykis.com/what_are_embeddings/ by @vicki.

#Llamafile is a Mozilla tool that packages a language model in a single executable file that will run on most platforms. It is 100% local and has been optimized to run on slower hardware, from Raspberry Pis to my 8yo laptop. It is based on llama.cpp which supports a plethora of models, not just LLMs: I chose all-minilm because it’s tiny (50MB) and has open code, research papers, and datasets. https://github.com/Mozilla-Ocho/llamafile https://huggingface.co/sentence-transformers/all-MiniLM-L6-v2

Второе пришествие мейнфреймов. Всё больше компаний хотят запускать ИИ у себя в офисе

Мейнфрейм IBM z16 во время лабораторных тестов в 2022 г, источник Приложения ИИ находят применение в бизнесе. Но есть проблема: корпоративные данные и документация представляют коммерческую тайну. Их нельзя передавать на сторону, тем более в облачную систему машинного обучения. Кроме того, что сама передача небезопасна, так ещё и публичная модель будет обучаться на наших секретах , а потом помогать конкурентам. В общем, у коммерческих компаний остаётся один вариант: поднимать собственный сервер или вычислительный кластер с ИИ. Таким образом, из эпохи облачных вычислений мы возвращаемся к старому доброму самохостингу, только сейчас это самохостинг GPU , серверы и мейнфреймы.

https://habr.com/ru/companies/ruvds/articles/867316/

#IBM #мейнфреймы #Telum_II #AnythingLLM #ChatGPT #Meta #обман #человечество #Mount_Diablo #IBM_z16 #самохостинг #ЦОД #датацентр #WatsonX #Copilot #Twinny #llamafile #Ollama #GPT4All #FraudGPT #WormGPT #Khoj #LocalAI #ruvds_статьи

Llamafile by Mozilla

Hi #fedifriends

has any of you experience with #llamafile by #Mozilla?
I would like to use a local #LLM #AI to get some help to create #XLST styles to run against #XLS and #CSV files with the goal of create #XML files without being an #XSLT champion...

I would like to do it locally, not depending stricly on #Linux and #Nvidia, with some #ethical software and looks like Llamafile check marks all those voices.

Thanks... 🙏

#fedihelp #fediask

Experimenting with #marimo, mastodon API, and #llamafile for my "Build Your Own Timeline Algorithm" project. 100% local, except those 11 seconds it took me to download my home timeline.

I have been pitching this idea around for almost two years now, and while I felt bad for not pushing more for it, I am glad that in the meantime new cool tech became available allowing me to do something even better than I imagined.

Seriously, #llamafile is amazing. On-device #llm with almost zero effort. Spectacular.

https://github.com/Mozilla-Ocho/llamafile

I also added Q4_K_M quantizations. #llm #llamafile

I also put some Llama 3.2 llamafiles onto the Docker hub: https://hub.docker.com/?namespace=maragudk . Quite nice for CI. #llm #llamafile

And if you don't want to build one yourself and just want to try out the small new Llama 3.2 models yourself, have some links:

- https://assets.maragu.dev/llm/Llama-3.2-1B-Instruct-Q8_0.llamafile
- https://assets.maragu.dev/llm/Llama-3.2-3B-Instruct-Q8_0.llamafile

#llm #llamafile

If you're into #llm stuff, I've created a small repository with scripts to create a #llamafile from a GGUF model.

In plain English, it makes it easy to download a language model and turn it into a file you can run directly. Super fun, super easy. https://github.com/maragudk/llamafile