LLM Reference

Ornith 1.5 Models by DeepReinforce

DeepReinforceMITOpen source
0 models

Last refreshed 2026-08-26. Next refresh: weekly.

Details

ResearcherDeepReinforce
LicenseMITOSI-approved
Commercial useCommercial use: permitted
Models0

About

Ornith 1.5 is Ornith/DeepReinforce's August 2026 open-weight family for Coding, Agents, and JSON / Tool use tasks. It extends the Ornith 1.0 self-scaffolding approach into a self-improvement loop that jointly generates training tasks, task-specific scaffolds, and solution rollouts. The family spans a 397B mixture-of-experts flagship, a 35B mixture-of-experts model, and a 9B dense model. The MIT-licensed 397B checkpoint is publicly downloadable on Hugging Face and documents a 262,144-token default context window plus OpenAI-compatible tool calling when self-hosted.

Decision facts

Best fit
codingagent workflowsstructured outputs
Capability starting point
Not tracked
Lowest tracked input
Not tracked
Closest related family
Ornith 1.0