The model is available on Hugging Face in FP8, NVFP4, GGUF, and MLX formats to support local deployment via Ollama, Hermes Agent, and AtomicChat. Osaurus integrated the model on launch day for Mac users without requiring cloud subscriptions, and AtomicChat published a quantization ladder that reduces memory requirements from 132.6GB to 35.0GB for use on 64GB Macs. The ecosystem also includes Unsloth for fine-tuning and RunInfra for API integration.
Tuesday, October 20, 2026
Story
Ornith Releases 122B Open Weight Model for Single NVIDIA DGX Card
AI9 hours ago
A new open weight large language model offers frontier level capabilities with an emphasis on exceptional inference efficiency. Released by Ornith as Ornith-1.5-122B, the system utilizes the MIT license and can execute on a single NVIDIA DGX card. The model scored 0.845 on the SWE-bench Verified benchmark, 0.813 on TB2.1, and 0.800 on CyberGym to demonstrate its proficiency in coding and agentic tasks.
Looking for more context or analysis?
Tilt chat provides reliable, tailored insights, alerts, and automations.