vb

Qwen released QvQ 72B OpenAI o1 like reasoning model on Hugging Face with Vision capabilities - beating GPT4o, Claude Sonnet 3.5 🔥

Model weights here: huggingface.co/Qwen/QVQ-72B...

BOOOOM! Meta released Llama 3.3 70B - 128K context, multilingual, enhanced tool calling, outperforms Llama 3.1 70B and comparable to Llama 405B 🔥 Comparable performance to 405B with 6x LESSER parameters ⚡

3.1 70B vs 3.3 70B: Code Generation > HumanEval: 80.5% → 88.4% (+7.9%) > MBPP EvalPlus: 86.0% → 87.6% (+1.6%) Steerability > IFEval: 87.5% → 92.1% (+4.6%) Reasoning & Math > GPQA Diamond (CoT): 48.0% → 50.5% (+2.5%) > MATH (CoT): 68.0% → 77.0% (+9%)

BOOOOM! Meta released Llama 3.3 70B - 128K context, multilingual, enhanced tool calling, outperforms Llama 3.1 70B and comparable to Llama 405B 🔥 Comparable performance to 405B with 6x LESSER parameters ⚡

Introducing Indic-Parler TTS - Trained on 10K hours of data, 938M params, supports 20 Indic languages, emotional synthesis, apache 2.0 licensed! 🔥 w/ fully customisable speech and voice personas! Try it out directly below or use the model weights as you want! 🇮🇳/acc

And.. here's a space to try out the model too: huggingface.co/spaces/ai4bh...

Introducing Indic-Parler TTS - Trained on 10K hours of data, 938M params, supports 20 Indic languages, emotional synthesis, apache 2.0 licensed! 🔥 w/ fully customisable speech and voice personas! Try it out directly below or use the model weights as you want! 🇮🇳/acc

you can just do things - ask AI to create your SQL queries and execute them right in your browser! 🔥 let your creativity guide you - powered by qwen 2.5 coder 32b ⚡ available on all 254,746 public datasets on the hub! go check it out today! 🤗

Fuck it! Structured Generation w/ SmolLM2 running in browser & WebGPU 🔥 Powered by MLC Web-LLM & XGrammar ⚡ Define a JSON schema, Input free text, get structured data right in your browser - profit!!

To showcase how much you can do with just a 1.7B LLM, you pass free text, define a schema of parsing the text into a GitHub issue (title, description, categories, tags, etc) - Let MLC & XGrammar do the rest! That's it, the code is super readable, try it out today! 🤗 huggingface.co/spaces/reach...

yo! nvidia finally released the weights for Hymba-1.5B - outperforms Qwen, and SmolLM2 w/ 6-12x less training trained ONLY on 1.5T tokens > massive reductions in KV cache size and improved throughput > combines Mamba and Attention in a hybrid parallel architecture with a 5:1 ratio and meta-tokens

Smol TTS keeps getting better! Introducing OuteTTS v0.2 - 500M parameters, multilingual with voice cloning! 🔥 > Multilingual - English, Chinese, Korean & Japanese > Cross platform inference w/ llama.cpp > Trained on 5 Billion audio tokens > Qwen 2.5 0.5B LLM backbone > Trained via HF GPU grants

Model weights on the hub, you can even run this on a Raspberry Pi! Go run, inference now! 🐐 huggingface.co/OuteAI/OuteT...

Massive week for Open Source AI/ ML Mistral Pixtral & Instruct Large - ~123B, 128K context, multilingual, json + function calling & open weights Allen AI Tülu 70B & 8B - competive with claude 3.5 haiku, beats all major open models like llama 3.1 70B, qwen 2.5 and nemotron

Jina AI Jina CLIP v2 - general purpose multilingual and multimodal (text & image) embedding model, 900M params, 512 x 512 resolution, matroyoshka representations (1024 to 64) Apple AIM v2 & CoreML MobileCLIP - large scale vision encoders outperform CLIP and SigLIP CoreML optimised MobileCLIP models

Massive week for Open Source AI/ ML Mistral Pixtral & Instruct Large - ~123B, 128K context, multilingual, json + function calling & open weights Allen AI Tülu 70B & 8B - competive with claude 3.5 haiku, beats all major open models like llama 3.1 70B, qwen 2.5 and nemotron

All the model checkpoints with URLs here: huggingface.co/posts/reach-...

Apple released blazingly fast CoreML models AND an iOS app to run them on iPhone! ⚡ > S0 matches OpenAI's ViT-B/16 in zero-shot performance but is 4.8x faster and 2.8x smaller > S2 outperforms SigLIP's ViT-B/16 in zero-shot accuracy, being 2.3x faster, 2.1x smaller, and trained with 3x fewer data

OH WOW! The Whale aka DeepSeek is BACK!! New model, with complete reasoning outputs and a gracious FREE TIER too! 🔥 Here's a quick snippet of it searching the web for the right documentation, creating the JS files plus the necessary HTML all whilst handling Auth too ⚡

I really hope they release a technical report or the model weights soon! 🤞

What a brilliant week in Open Source!

Smol TTS models are here! OuteTTS-0.1-350M - Zero shot voice cloning, built on LLaMa architecture, CC-BY license! 🔥 > Pure language modeling approach to TTS > Zero-shot voice cloning > LLaMa architecture w/ Audio tokens (WavTokenizer) > BONUS: Works on-device w/ llama.cpp ⚡

Three-step approach to TTS: > Audio tokenization using WavTokenizer (75 tok per second) > CTC forced alignment for word-to-audio token mapping > Structured prompt creation w/ transcription, duration, audio tokens Check out the model here: huggingface.co/OuteAI/OuteT...