Open Source
The best open-source projects, tools, and communities
NVIDIA GR00T N1.7: The Open Robot Brain Trained on Human Video
NVIDIA GR00T N1.7 is a 3B-parameter open VLA model pre-trained on 20,854 hours of human egocentric video — and it has a dexterity scaling law.
By Aisha Patel · 6 min · May 5, 2026
Postiz: The 29.6K-Star Open-Source Social Scheduler Killing Buffer
Postiz hit 29.6k stars and #1 on Product Hunt with a self-hostable, AGPL-3.0 social media scheduler — and a new agent CLI that lets Claude post for you.
By Marcus Rivera · 6 min · May 4, 2026
VibeVoice: Microsoft's Open-Source Frontier Voice AI Hits 33K Stars
Microsoft's VibeVoice family delivers 60-minute single-pass ASR, sub-300ms streaming TTS, and an unusual content-safety story—all under MIT license.
By Marcus Rivera · 7 min · May 2, 2026
Trinity-Large-Thinking: 400B U.S.-Made Open Reasoning Model
Arcee AI just released Trinity-Large-Thinking, a 400B-parameter sparse MoE open-weights model trained on 2,048 NVIDIA B300 GPUs for $20M. Apache 2.0, agent-tuned, and 96% cheaper than Claude Opus 4.6.
By Aisha Patel · 7 min · Apr 30, 2026
Nemotron 3 Nano Omni: NVIDIA's 30B Open Model Sees and Hears
NVIDIA's Nemotron 3 Nano Omni replaces the Frankenstein perception stack with one 30B-parameter open model. Audio, video, documents, and text — single forward pass, 9x the throughput.
By Marcus Rivera · 6 min · Apr 29, 2026
Archon OS: The Open-Source Brain That Makes Claude Code Remember
Archon OS by coleam00 is a 13.7K-star open-source MCP server that gives AI coding assistants persistent knowledge and task management.
By Marcus Rivera · 6 min · Apr 21, 2026
Goose: Block's Open-Source Local-First AI Agent Hits 35K
Goose, Block's local-first AI agent, just passed 35,300 GitHub stars and joined the Linux Foundation's Agentic AI Foundation as a founding project.
By Marcus Rivera · 5 min · Apr 20, 2026
Voicebox: The Local-First Voice Cloning Studio for Mac and Windows
Voicebox is an MIT-licensed local voice-cloning studio on Tauri, powered by Qwen3-TTS, with a full REST API and an MLX backend that runs 4–5x faster on Apple Silicon.
By Marcus Rivera · 7 min · Apr 18, 2026
NVIDIA Ising: Open-Source AI Models That Make Quantum Computing Actually Work
NVIDIA released Ising on April 15, 2026 — a family of open-source AI models that deliver up to 2.5x faster and 3x more accurate quantum error correction. The stack ships on GitHub and Hugging Face, integrates with CUDA-Q and NVQLink, and is already running in labs at Harvard, Fermilab, and IQM.
By Marcus Rivera · 6 min · Apr 17, 2026
GLM-5.1: The Open-Source 754B Model That Works for Eight Hours Straight
Z.ai releases GLM-5.1 under MIT license — a 754B MoE model that tops SWE-Bench Pro and runs autonomously for up to 8 hours. Open-source agents just got serious.
By Marcus Rivera · 5 min · Apr 15, 2026
Ghost Pepper: 100% Local Speech-to-Text for macOS
Ghost Pepper is an open-source macOS app that runs speech-to-text and meeting transcription entirely on Apple Silicon. No cloud, no subscriptions, no data leaving your machine.
By Marcus Rivera · 5 min · Apr 15, 2026
Open SourceRay: The Open-Source AI Financial Advisor That Runs on Your Laptop
Ray is an MIT-licensed terminal app that plugs into 12,000+ banks via Plaid and answers questions about your money with no cloud — your data never leaves your disk.
By Marcus Rivera · 4 min · Apr 12, 2026
Hermes Agent: The Open-Source AI Agent That Learns How You Work
NousResearch's Hermes Agent just hit 60K GitHub stars with a built-in learning loop, persistent memory, and autonomous skill creation. Here's what makes it different.
By Marcus Rivera · 4 min · Apr 12, 2026
Cohere Transcribe: The Open-Source ASR Model That Dethroned Whisper
Cohere Transcribe is a 2B-parameter open-source speech recognition model that tops the HuggingFace ASR Leaderboard with 5.42% WER — 28% better than Whisper Large v3, and 3x faster.
By Marcus Rivera · 5 min · Apr 11, 2026
Gemini CLI: Google's Open-Source Terminal Agent Hits 101K GitHub Stars
Google's open-source terminal AI agent crosses 101K GitHub stars with v0.36.0, bringing Gemini 3 models, 1M-token context, MCP support, and strict sandboxing to the command line.
By Marcus Rivera · 4 min · Apr 9, 2026
Google Gemma 4: Four Open Models That Punch Above Their Weight
Google DeepMind releases Gemma 4 — four open-weight models from 2B to 31B parameters that rank #3 globally, run on consumer hardware, and ship under Apache 2.0.
By Marcus Rivera · 4 min · Apr 8, 2026
Moondream 3: The 9B Vision Model That Runs Like a 2B
Moondream 3's MoE architecture delivers 9B-class vision understanding with 2B inference costs and state-of-the-art segmentation.
By Marcus Rivera · 4 min · Apr 1, 2026
Voxtral TTS: Mistral's Open-Weight Speech Model Challenges ElevenLabs
Voxtral TTS is Mistral's first open-weight speech model — 4B parameters, 9 languages, and ElevenLabs-tier quality you can run locally.
By Marcus Rivera · 4 min · Mar 31, 2026
Mistral Small 4: One Open-Source Model Replaces Three Separate AI Products
Mistral Small 4 unifies instruct, reasoning, and vision in one 119B MoE model under Apache 2.0.
By Marcus Rivera · 4 min · Mar 30, 2026
Qwen 3.5 Small: Alibaba's 9B Model That Beats GPT-OSS-120B
Alibaba's Qwen 3.5 Small series delivers natively multimodal AI models from 0.8B to 9B parameters under Apache 2.0, with the 9B scoring 81.7 on GPQA Diamond — beating OpenAI's gpt-oss-120B.
By Sarah Chen · 5 min · Mar 29, 2026
Biome v2.4: The Rust-Powered Toolchain Replacing ESLint and Prettier
Biome v2.4 is a Rust-powered toolchain replacing ESLint and Prettier with 35x faster formatting, 465 lint rules, and adoption by major tech companies.
By Marcus Rivera · 6 min · Mar 29, 2026
OpenClaw: The Self-Hosted AI Agent That Hit 247K GitHub Stars
OpenClaw turns your existing chat apps into a unified self-hosted AI agent — 247K GitHub stars in four months, and a bet that the best AI interface is the one you already have.
By Marcus Rivera · 6 min · Mar 29, 2026