Ornith (35B,9B) + Hermes,Zed: THE FULLY PRIVATE LOCAL AGENT is ACTUALLY HERE!

In this video, I'll be telling you about Ornith, a new family of open agentic coding models from DeepReinforce. These models range from a small 9B model to a huge 397B model, and they focus on coding, tool use, and agentic workflows. I’ll also share my thoughts after testing the 35B model locally. -- Key Takeaways: 🚀 Ornith is a new open family of agentic coding models from DeepReinforce. 🧠 The models are built on top of Gemma and Qwen, with a focus on coding, tool use, and agentic workflows. 🛠️ Ornith claims to learn its own scaffold, including planning, retries, tool use, file checks, and testing. 📊 The 397B model shows very strong benchmark results, but it is not practical for most local users. 💻 The 35B and 9B models are the most interesting options for local machines and servers. 🔓 The models are open, MIT licensed, and available in GGUF versions for easier local usage. ⚙️ Proper setup matters because Ornith can produce reasoning tags and tool-call blocks. 👍 In testing, the 35B model worked well with agents, tool calls, web search, and local workflows.