Build Anything with Ornith-1.5: Here's How!

Ornith 1.5: Business-First Guide to the New Open-Source AI

The video is presented by the digital avatar of Julian Goldie, CEO of Goldie Agency, and focuses on how business owners and agencies can actually use Ornith 1.5, a free, MIT-licensed AI model that rose to the top of Hacker News shortly after release. It intentionally avoids pure benchmark talk, instead demonstrating real workflows that generate content, landing pages, onboarding sequences, and ad campaigns.

What Makes Ornith 1.5 Stand Out 🧠

• Ornith 1.5 generates its own tasks: it identifies weaknesses, creates a test for them, attempts the test, scores itself, and repeats the process.
• It deliberately targets a ~20% success rate so challenges are difficult enough to generate meaningful learning.
• The system loop is simple: task → scaffold → attempt → score → harder task.
• Ornith’s published number on Terminal-Bench 2.1 is 86.1, beating Claude Opus 4.8’s 85.0. These are self-reported and not yet independently reproduced, so the data should be treated as the vendor’s claim for the time being.

Which Model to Use

• 9B tier: Lightweight, includes a ~1.5GB mobile version for iPhone/Android; runs privately on-device with no cloud or subscription.
• 35B tier: Uses only ~3B active parameters per token, making it compute-efficient and powerful enough for private-server deployment. This is described as the sweet spot for most developers and businesses.
• 397B tier: The flagship model, designed for enterprise-grade infrastructure with around 8×H200 GPUs; suitable for agencies and large cloud-based operations.

Business Workflows Demonstrated

1. Content repurposing: Feed one coaching-call transcript into the model and generate five short-form hooks, three email subject lines with preview text, ten conversational posts for business owners, and a 14-day content calendar in under 60 seconds. One call becomes nearly a full month of reach.
2. Landing page creation: From a blank page, Ornith 1.5 writes a full page with headline, sub-headline, three benefit sections, objection handling for people who fear AI is too complex, and a strong call-to-action.
3. Email onboarding: A five-email sequence is generated for new community members — welcome, member map access, 30-day roadmap, coaching call booking, and a day-14 check-in.
4. Ad campaigns: With a direct-response marketing angle, the model produces three ads targeting distinct pain points: too much admin, not enough customers, fear of falling behind with AI. Each includes hooks, ad body copy, CTAs, plus A/B test variations for the headline.

Technical Integration & Setup 💻

• Requires up-to-date runtimes: Transformers 5.8.1, vLLM 0.19.1, or SGLang 0.5.9 or higher.
• The easiest start is to run Ornith locally via vLLM, exposed as an OpenAI-compatible endpoint — so your current OpenAI-based tools can connect to it by changing one line of configuration.
• The 9B model can also be run easily via Ollama, while Apple users can leverage the official MLX collection on Hugging Face.
• On-device usage is still early, but the gap between local computing vs. cloud is closing quickly.

Extra Context From the Video ⏳

The creator also promotes the AI Profit Boardroom, a community of over 3,700 business owners using AI to automate content, lead generation, client reporting, and other workflows. He also highlights the AI Success Lab with 87,000 members and 100+ free AI case study templates, although the emphasis of the video remains on practicing operational use of the AI model itself.

Key Takeaway

Ornith 1.5’s main business value is not simply its benchmark score, but the ability to run a private, self-improving model that just plugs into the same tools you’re already using. The 35B model is the recommended starting point for practical business implementation. Since the reported performance claims are not yet independently verified, teams should benchmark the model further before critical workloads. Still, the strategies, templates, and deployment paths shown in the video give early adopters a clear reason to start experimenting now rather than waiting for the market to catch up.