Archive, page 5
Claude Haiku 5.5, Anthropic's smaller model, reaches GitHub Copilot, with pricing and best-use guidance published. The source text does not give full pricing or context details.
Takes for Mac is a Mac app for recording talking-head videos, where Claude Code handles the editing. Launched on Product Hunt with discussion and a link.
SupoClip is an open-source AI clipper that cuts long videos into short viral clips. Launched on Product Hunt with a link and discussion.
Version 0.25.1 of Qwen Code, an AI-powered coding assistant CLI, installable via npm i -g @qwen-code/qwen-code.
LiteLLM is a gateway that calls 100+ LLM APIs in OpenAI format with cost tracking and load balancing. v1.105.0 signs all Docker images with cosign.
Version 1.19.0 of OpenCode, a terminal-based AI coding agent, installable via npm i -g opencode-ai.
Open WebUI is a self-hosted chat interface for Ollama and OpenAI-style APIs. v0.12.0 adds realtime voice calls using an OpenAI Realtime voice model.
Aurelio-finance is an open-source Python app to track net worth, investments, ETFs, cash and debts locally, with an AI financial advisor. It gained 2.1k stars in its first week.
Openchart is an open-source TypeScript trading workspace that runs agents for trading tasks. It gained 151 stars in its first week.
Absent-author is a Python toolkit with an evidence list and agent skills to help identify research papers produced by AI without a human author standing behind them. It gained 122 stars in its first week.
Paperfold is a Python tool that lets users zoom into any research paper to understand it better. It gained 119 stars in its first week on GitHub.
Mortiflix-oss is a JavaScript studio you run locally: Claude builds a motion video step by step and you approve each stage. Bring your own Claude Code or API key. It gained 526 stars in its first week.
Bops is a TypeScript project for running business operations with a team of AI bots, each with its own computer, email and phone number. It gained 243 stars in its first week.
P3-stack is a shell-based toolkit for T3 Code that adds delegated subagents, worktree threads, PR watching and scheduled runs. It gained 143 stars in its first week.
A static Hugging Face Space shows a Decision Index leaderboard and a news tab tracking open reproductions of TypeSafe's Jev, including decoding and diffusion approaches. It has 574 hearts.
A Gradio web demo where you can generate images from text or edit an uploaded image, running the Qwen-Image-2.1 Uncensored GGUF model on ZeroGPU. Hosted demo, no install needed.
A web tool where you look up any Hugging Face model to see daily download history, likes, and fine-tunes, quantizations and merges built on it. Compare up to five models on one chart.
Two demo apps swap heads or faces in images using the BFS LoRA on Qwen-Image-2.1, with an optional 6-step turbo mode. Non-commercial license terms apply to the base model and turbo LoRA.
A distilled version of Qwen-Image-2.1 that generates and edits images in 6 steps without classifier-free guidance, about 5x faster than the 40-step base model. Demo Space available.
An image model for text-to-image generation and image editing, with native RGBA transparency support. Interactive demo Space is available on Hugging Face, with weights and blog linked.
A Gradio app that turns an uploaded start frame and a prompt into a short clip using Wan 2.2 Lightning, with selectable KV adapters loaded from a private model repo.
A browser demo where you play chess against the complete connectome of an adult fruit fly, using 138,639 neurons, running on WebGPU. Board drives visual neurons and the brain outputs a move.
An experimental web app for editing images with FLUX.2 Klein 9B and multiple LoRAs. Uploaded images are deleted after 7 days, and the author notes uploads may be logged.
A Gradio demo for generating and editing images with the Qwen-Image-2.1 Uncensored GGUF model, using its recommended Q4_K_M quantized checkpoint on ZeroGPU.
Open Apache 2.0 models that take a state (text, JSON, images or video) and typed questions, returning probabilities for every option in one forward pass. Clef-omni is on Hugging Face.
An open Apache 2.0 mixture-of-experts model focused on German and English, with explicit reasoning mode, tool calling and long-context efficiency. Weights are on Hugging Face.
A tiny open Turkish text-to-speech model, about 34 MB in total, with a 5.6M-parameter DiT acoustic model and 3M vocoder. Reports 0.92% WER on Freya-TR-Eval and very fast speech output.
Open decision models that take a state and questions and return typed answers in one forward pass. d1-3B is built on LFM2.5-VL-3B and claims the top score under 10B on the Decision Index.
Quantized GGUF versions of EmbeddingGemma 2, an open Apache 2.0 embedding model that maps text, code, images, video and audio into one space. Unsloth provides run guides.
A community fine-tune of Qwen3.8-27B, post-trained with several rounds of SFT and RLOO and MTP/DFlash variants, available as FP8, NVFP4, INT8 and GGUF weights.
A quantized GGUF version of Qwen3.8-Flash-Next where the refusal direction was removed by swapping 144 tensors for abliterated weights. Context is 262K tokens, under Apache 2.0.
A LoRA for Qwen-Image-2.1 that dresses a person in a supplied outfit image while keeping the original picture frame, face and background from shifting.
A 3-billion-parameter model that generates images directly in pixel space without a VAE. Fine-tuned, it estimates depth and restores and upscales images. Apache 2.0 licensed.
A 3.0 bpw EXL3 quantization of a weight-edited GLM-5.3 variant, with 753B total parameters, for running locally. Not affiliated with Z.ai, and MIT licensed.
An Apache 2.0 27B decision model built on Qwen3.8-27B that lifts the Decision Index score from 56.4 to 61.56, beating Jev by over 3.5 points using the same backbone.
A JavaScript repo that lets any AI agent call Codex's built-in image generation through MCP, a CLI, or a skill, with prompt tips for Xiaohongshu covers and posters. Gained 127 stars in its first week.
A dataset of 5.6 billion TikTok video records under CC-BY-NC 4.0, with the scraper source code and a guide to TikTok's mobile API, including endpoints and request signing.
Open RL training environments for LLM agents across software engineering, cyber vulnerability reproduction, knowledge work, web development and music, with executable tests, Docker images and training code.
A snapshot of arXiv metadata, version history, submission files and rendered documents covering 3,148,796 papers, with a PDF for 99.47% of them, plus SHA-256 digests.
A 1K to 10K-row Apache 2.0 dataset of doctor-patient ChatML conversations covering every human disease on a list, generated by Claude Opus 5.5, one agent per disease. Sequel to an Opus 4.8 version.