Loading activity…
@mervenoyann
(mer-veh) open-sourceress at @huggingface 🧙🏻♀️ DM me for any feedback about HF 🤗 https://github.com/merveenoyan/smol-vision
AI
A Nvidia GTC attendee in Berlin said they started using GPUs for deep learning as a CIFAR professor in 2009.
AI
The broadcast description promises “massive speed-ups” with speculative decoding; a post announced the stream was live.
AI
Hugging Face shared a “Local AI 202” broadcast promising massive speed-ups with speculative decoding.
AI
A user says the option to “vibe code” apps, use community creations or jailbreak the r1 has renewed their interest.
AI
Artificial Analysis scores it 39 on its Intelligence Index, the highest score for a US open-weights model in that ranking.
Technology
A participant suggests Llama for beginners; its site says it enables speculative decoding automatically on supported models.
AI
A user shares an illustration of the difference, saying they think they’ve finally grasped the distinction.
AI
Hugging Face shared the event; one speaker points beginners to Llama for automatic speculative decoding.
AI
Reflection says the release includes Mirror CLI, its open-source training harness, and collaborations with 40-plus partners.
Technology
Meta says its 816B-parameter model can run locally across two M5 Ultra Mac Studios.
AI
finkd says the model runs well on new Mac Studios, crediting work with Apple and John Ternus.
AI
PrimeIntellect says it built 2,500 hard enterprise tasks with NVIDIA, modeled on AutomationBench.
AI
Nvidia AI said on October 14 that it had a model coming, noting many others launching that week.
AI
A user praises what people are building with EmbeddingGemma 2 and welcomes the attention embedding models are getting.
AI
A user says seeing how few env tasks they have prompts repeated generation, making a training deadline necessary.
AI
Its author says it’s the first conceptual guide aimed at getting more performance from a local setup.
AI
Its creators say the platform trains Qwen-3.8-27B on agents’ environments and adds evaluation scores to a leaderboard.
AI
Its creator says it covers prefill versus decode, MoE versus dense models, and VRAM versus unified memory.
Technology
The creator says it also covers MoE versus dense models and VRAM versus unified memory, and welcomes reuse with attribution.
AI
Mistral says Mistral Large 4 is live via API in public preview now, with open weights due at the end of October. Reuters says the launch is part of Mistral’s push to narrow the gap with leading Chinese open-weight models.
