DeepReinforce Releases Ornith 1.0, a 35B Reasoning Model

DeepReinforce AI has published Ornith 1.0, a 35-billion-parameter language model geared toward conversation and reasoning tasks. The release arrives on Hugging Face in GGUF format, the packaging most commonly used for running models locally through runtimes like llama.cpp and Ollama.

At 35B parameters, Ornith 1.0 is a dense model rather than a mixture-of-experts design, placing it in a middle tier that balances capability against the hardware needed to run it. That size is large enough to handle multi-step reasoning but still within reach of well-equipped workstations, especially with the quantized weights that GGUF makes straightforward to distribute.

Why it matters

The permissive MIT license is the most notable detail here. It allows commercial use, modification, and redistribution with minimal restrictions — a friendlier posture than the custom community licenses attached to many open-weight releases.

• Dense 35B architecture, listed for text and reasoning workloads
• Distributed as GGUF for local and self-hosted setups
• MIT license, permitting commercial use and modification

As an initial 1.0 release from a less-established name, Ornith enters a crowded field where models in this size class are judged on real-world performance rather than promises. DeepReinforce has not published detailed benchmark claims alongside the weights, so the practical test will come as the community puts the model through its paces. The weights are available now on Hugging Face.