The Local AI Architecture Upgrade: Ollama's Jev-Style Decision Models

ew episode 🎧 The Local AI Architecture Upgrade Ollama 0.35 brings Jev-style decision models to your local machine. Instead of generating paragraphs, these small models make fast, strictly typed decisions and return clean JSON. In this episode: • What a Jev-style model is, and how the /v1/systemone endpoint works • Nimble 9B (Bespoke Labs) and the Tev1 4B and 0.8B models (Together AI) • ~91 ms per decision, fully local, with no cloud API costs • Plugging it into C# .NET and WPF apps over a simple REST endpoint • Using a tiny 0.8B model as a router in front of heavier models like Gemma 4 or Kimi Try it: update to Ollama 0.35 and run `ollama pull nimble`. How would you use structured decision models in your projects: data triage, or a router for your bigger LLMs? Let me know! Narration voice and background music are AI-generated. Thanks for the coffee ☕ It keeps these episodes coming.