Some nice quantization work makes our Gemma 3 model even more compact.
OPEN SOURCE
-
Drupal Developer Survey Now Open for Community Input
By
–
The @drupal Developer's Survey takes less than 10 minutes and is available for 2 more days! Take it if you haven't yet –> https://
bit.ly/4jrXn6i #drupal #foss -
Meta-FAIR Releases Open Source AI Model
By
–
Open source AI FTW.
— Yann LeCun (@ylecun) 18 avril 2025
From Meta-FAIR. https://t.co/D9PrGYzeeoOpen source AI FTW.
From Meta-FAIR. -
InternVL3 dethrones Qwen 2.5 as new king of open VLMs
By
–
New king of open VLMs: InternVL3 takes Qwen 2.5's crown! -> InternVL have been a wildly successful series of model : and the latest iteration has just taken back their crown thanks to their superior, natively multimodal vision training pipeline. Most of the vision language
-
FramePack: 30 FPS Video Generation on 6GB Laptop GPU
By
–
FramePack can diffuse thousands of frames at full 30 FPS using 13B models — all on just 6GB of laptop GPU memory! And the results? Surprisingly good.
— 机器之心 JIQIZHIXIN (@jiqizhixin) 18 avril 2025
Project: https://t.co/sueGQRQ0Mq
Paper:https://t.co/hAY99BgMzx
Code: https://t.co/G85y03ic4T pic.twitter.com/AjgqrG0xGxFramePack can diffuse thousands of frames at full 30 FPS using 13B models — all on just 6GB of laptop GPU memory! And the results? Surprisingly good.
Project: https://
lllyasviel.github.io/frame_pack_git
page/
…
Paper: https://
lllyasviel.github.io/frame_pack_git
page/pack.pdf
… Code: https://
github.com/lllyasviel/Fra
mePack
… -
Codex CLI Updates: MCP Support and Plugin System Coming
By
–
Amazing to see the community excitement on Codex CLI! Still early days: we're adding support for MCP, local/different provider models, and a native plugin system. In the immediate term, also fixing rate limit issues that some people have reported. Keep the feedback coming!
-
LiteLLM Now Natively Supports Inference Providers
By
–
ICYMI, Inference Providers are supported natively in LiteLLM now!
-
Remote HTTP servers and OAuth support for AI clients
By
–
More servers w/ the new remote HTTP support and OAuth would be nice. More usable clients; we're building one @Lutra_AI
-
Llama 4 Open-Source: Leaked Data Concerns and Meta Transparency
By
–
Llama 4 – open-source, but you can't download the right version? Some leaks claim that the team snuck benchmark test-set data into training. We can't really say what's true and what's not, but Meta surely didn't tell us everything here. But that's just to say that we can't