# Sources for Ornith-1.0 Local Agentic Coding Model Guide

Generated: 2026-07-03

## Primary video

- YouTube Short: “Ornith 1.0 Beats Claude Opus 4.7 — and It's Free #aimodel #coding” by DIY Smart Code
- Video ID: `8Cw1jX6o9w0`
- URL: https://www.youtube.com/watch?v=8Cw1jX6o9w0
- Upload date from yt-dlp metadata: 2026-06-28
- Runtime from yt-dlp metadata: 133 seconds
- Local metadata/transcript folder: `/Users/alanwheat/.hermes/guide-publishing/youtube-8Cw1jX6o9w0/`

## Project / model links from the video description

- DeepReinforce blog: https://deep-reinforce.com/ornith_1_0.html
  - Verified HTTP 200 on 2026-07-03.
  - Page title captured: “Ornith-1.0: Self-Scaffolding LLMs for Agentic Coding | DeepReinforce Blog | Jun. 2026”.
  - Key details used: self-scaffolding RL framing, model sizes, and benchmark claims.

- Ollama library page: https://ollama.com/library/ornith
  - Verified HTTP 200 on 2026-07-03.
  - Page title captured: “ornith”.
  - Key details used: `ollama run ornith`, `ornith:9b`, `ornith:35b`, 5.6GB 9B model entry, 256K context listing, MIT license note in readme.

- Hugging Face collection: https://huggingface.co/collections/deepreinforce-ai/ornith-10
  - Verified HTTP 200 on 2026-07-03.
  - Page title captured: “Ornith-1.0 - a deepreinforce-ai Collection”.
  - Key details used: available model repositories including 9B, 35B, 397B, GGUF, and FP8 variants.

- X / Twitter account from video description: https://x.com/ornith_
  - Verified HTTP 200 on 2026-07-03.
  - Page title captured: “Ornith (@ornith_) / X”.

## Notes on transcript accuracy

The auto-caption transcript misheard “Ornith” as “Open Earth” in several places. The guide uses “Ornith” because that spelling is confirmed by the video title, description, Ollama page, Hugging Face collection, and DeepReinforce blog.

## Main claims carried into the guide

- Ornith-1.0 is presented as an open-source/open-weights family for agentic coding.
- Claimed sizes include 9B Dense, 31B Dense, 35B MoE / 35B public variant, and 397B MoE / flagship; public pages verified 9B, 35B, and 397B repositories plus GGUF/FP8 variants.
- Claimed flagship benchmark numbers from the source material include 77.5 on Terminal-Bench 2.1 and 82.4 on SWE-bench Verified.
- Ollama currently lists `ornith:9b` at 5.6GB and `ornith:35b` at 21GB, with `ollama run ornith` as the default command.
- The guide treats benchmark claims as vendor/project claims and recommends local validation before replacing a daily coding model.