The Information Machine

Ornith-1.0: Self-Scaffolding LLMs for Agentic Coding

Simon Willison · Simon Willison · 2026-06-29

DeepReinforce releases Ornith-1.0, an MIT-licensed open-weights model family spanning 9B to 397B parameters built on Gemma 4 and Qwen 3.5, claiming state-of-the-art coding benchmark performance among comparable open-source models.

Open original ↗

Appears in

Extraction

Topics: open-source-llmsagentic-codingmodel-releaselocal-llmscoding-benchmarks

Claims

  • Ornith-1.0 is an MIT-licensed open-weights model family with variants including 9B Dense, 31B Dense, 35B MoE, and 397B MoE.
  • The models are built on Gemma 4 and Qwen 3.5, both Apache 2.0 licensed, making them compatible with open-source distribution.
  • Ornith-1.0 achieves state-of-the-art performance among open-source models of comparable size on coding benchmarks.
  • Early testing shows the 35B MoE GGUF runs the agent harness across many tool calls proficiently at 103 tokens per second on local hardware.

Key quotes

Initial impressions are very good - it seems to be able to run the agent harness over many tool calls in a proficient way.
Gemma 4 is Apache 2.0 licensed (and not bound by the janky additional Gemma Terms of Use that afflicted the previous Gemma models)