C4AI Command A 111B

🇨🇳 Sources: DeepSeek is speeding up the release of its R2 AI model, which was originally slated for May, but the company is now working to launch it sooner.

LlamaCon on April 29: Meta to share the latest on Open Source AI developments

Zonos-v0.1 beta by Zyphra, featuring two expressive and real-time text-to-speech (TTS) models with high-fidelity voice cloning. 1.6B transformer and 1.6B hybrid under an Apache 2.0 license.

Mistral’s new “Flash Answers”

Anthropic CEO is coping and seething over DeepSeek

R1+Sonnet set a new SOTA on the aider polyglot benchmark, at 14X less cost compared to o1

Llama 4 is going to be SOTA

Comparison: Llama 3.2 vs Gemma 2 vs Mistral

Why does 1206 love Bengali so much?

FishSpeech v1.5 - multilingual, zero-shot instant voice cloning, low-latency Only 500M params - #2 ranked on TTS-Arena

KoboldCpp 1.79 - Now with Shared Multiplayer, Ollama API emulation, ComfyUI API emulation, and speculative decoding

Tülu 3 -- a set of state-of-the-art instruct models with fully open data, eval code, and training algorithms

Llama 4 Models are Training on a Cluster Bigger Than 100K H100’s: Launching early 2025 with new modalities, stronger reasoning & much faster

Cohere releases Aya Expanse multilingual AI model family

IBM Granite 3.0 Models

Is it possible to achieve very long (100,000+) token outputs?

NVIDIA's latest model, Llama-3.1-Nemotron-70B is now available on HuggingChat!

Benchmark Your LLM Against Korea’s Most Challenging Exam!

Is it possible to run some simple LLM (e.g. llama2) using very low amounts of RAM (e.g. 16MB)?

A1C 5.4 but Fasting 120, 2-Hour Post-Glucose 190 – How Concerned Should I Be?

F5-TTS: A Fairytaler that Fakes Fluent and Faithful Speech with Flow Matching [Best OS TTS Yet!]

Zamba 2 2.7B & 1.2B Instruct - Mamba 2 based & Apache 2.0 licensed - beats Gemma 2 2.6B & Mistral 7B Instruct-v0.1

Local LLama 3.2 on iPhone 13

OLMoE 7B is fast on low-end GPU and CPU