Loading...

Tag trends are in beta. Feedback? Thoughts? Email me at [email protected]

Alleged Mythos Class Anthropic Checkpoint Leak

Llama.cpp v1 Is Out

Nvidia/DeepSeek-V4-Pro-0813-NVFP4

Show HN: DeepSeek-Coder-V2-Lite Under 5GB: 65 TPS 100% In-VRAM on 6GB GPU

Kimi K3.1 Pro model weights are out

Yang-Mills, OpenAI proof, and the mass gap problem

Show HN: Nano-Edge-TTS – 70x realtime, Lightweight Kokoro-based TTS model

The MoE Training Handbook

0.5B base model trained on 60B tokens rivals Qwen3.5-2B trained on 30T+ tokens

DeepSeek V4 Pro 0813

New Nvidia Cosmos 3 Open Models for Interactive World Simulation

Nvidia Nemotron 3.5 Transcribe: open ASR for 40 languages

DeepSeek-v4.1-Flash-1017

Show HN: Hurston-1 open source TTS that speak Southern American and Afro-English

Hy-Nomas, Tencent's new open-source world model you can steer with text

Unlimited-OCR-2: Baidu open-sources an OCR model for unlimited-horizon parsing

Welcome Beam: a 501B open model you can own, customize, and scale

Flux 3 Image Klein

Fish Audio Drama 3: an open-weight, directable TTS model

Show HN: TerrainSR – fast, realistic heightmap upscaling model

LiquidAI D1 Decision Model

LensVLM-9B by Apple

Show HN: Mahabharata Characters – 307 figures with gunas and decision primitives

Flux 3 Action: A 7B open-weight world action model for robots

NASA and IBM Open Source Lunar Mapping Tools

LensVLM: Compressing long context as images, expanding only relevant pages

Show HN: Swift-Qwen3.8-27B, -58.3% thinking, x1.95 speed, accuracy of xhigh

(Tech Report) DeepSeek-v4.1-Flash: Pushing Limits of KV Cache Compression [pdf]

DeepSeek-v4.1-Exp

HuggingFace: Security.txt

More →