Ornith-1.5-9B

Ornith-1.5-9B

Handles coding and terminal tasks as a 9B dense language model sized for a single GPU. For developers who want a small open model for software work, with a mobile variant for phones.

Made by ornith-ai on Hugging Face. These notes are Mindket's, written from its model page and model card; the model's own files and license are on Hugging Face.

About

What it is: Ornith-1.5-9B is the lightest member of the Ornith-1.5 family. It is a dense 9B model aimed at running on one GPU. The family is trained with a self-improvement loop that generates its own training tasks and learns from the results with reinforcement learning. A quantized Ornith-1.5-9B-Mobile variant targets phones.

Who it suits: Developers who want a small open model for coding and terminal work, without a large GPU cluster.

Industry | Who uses it | How they use it | Example
Software | Developers | Run a small coding model on one machine | Fix a bug in a local project
DevOps | Engineers | Try a model on terminal-style tasks | Script a routine server chore
Research | AI researchers | Study a model trained by self-improvement | Compare it with the Qwen3.5 and Gemma models in the card

If you need the strongest results on every task, the card's larger family members and bigger models are worth a look first.

Where to use it:

• Your own computer with Python, on Windows, macOS or Linux
• A single GPU, which the card names as the target for this size
• Mobile devices, through the separate quantized Mobile variant

Use case | Who | What they get back
Coding help | Developers | Code and fixes written by a small model
Terminal tasks | Engineers | Command-line work attempted by the model
On-device tests | App builders | A phone-sized quantized variant to try

How to set it up:

• Python with pip
• A GPU with room for the weights, about 19.3 GB in BF16, or about 4.8 GB at 4-bit (an estimate)
• The transformers library, which the model is made for

1. Install the Hugging Face hub tool.
2. Download the model files.
3. Load them with the transformers library.

pip install -U huggingface_hub
hf download ornith-ai/Ornith-1.5-9B

How to use it: Load the model with transformers and give it a coding or terminal task as text. The part of the card shown here has no run snippet, so check the rest of the model card for the exact way to call it.

Example: You give it a description of a bug in a small project and ask for a fix. The model replies with a proposed code change as text. This illustrates the kind of coding task its benchmarks measure and is not a tested run.

For newcomers and experts: If you are new, start with a small coding question and see how the answer reads. Check your GPU memory first, since the full-precision weights are about 19.3 GB.

For experts, Ornith-1.5 follows Ornith-1.0, which started from Qwen3.5 and Gemma4 and received more training stages on top. The 1.5 loop optimizes task generation, scaffold construction and solution attempts together, instead of using fixed human-made tasks.

The model generates new training tasks
It finds strategies for solving them
Reinforcement learning updates the policy
The loop repeats with harder tasks

Good to know: The card compares scores on coding benchmarks. Terminal-Bench 2.1 is shown here for the Terminus-2 setup.

Title: Terminal-Bench 2.1 (Terminus-2) scores from the card
Qwen3.6-35B-A3B: 52.5 points
Ornith-1.5-9B: 46.2 points
Ornith-1.0-9B: 43.1 points
Gemma-4-31B: 42.1 points
Qwen3.5-9B: 21.3 points

• License: MIT, as the Hugging Face record lists it.
• Cost to run: no API key or account is named for a downloaded copy. You supply the GPU or other hardware.
• Data it reaches: downloading reaches Hugging Face. The card does not describe the model reaching files, the web or an account.
• Not for: the card names no banned uses. The larger Qwen3.6-35B-A3B scores higher on the coding benchmarks it lists.

On Hugging Face, 13 November 2026

• Task: text generation
• Library: transformers
• Parameters: 9.7B
• Weights: about 19.3 GB in BF16; about 4.8 GB at 4-bit (estimate)
• Formats: Safetensors
• Downloads in the last 30 days: 534,050
• Likes: 414
• License: mit
• Quantized versions: 102
• Fine-tunes: 44
• Tags: transformers, safetensors, qwen3_5, image-text-to-text, text-generation, conversational

Links

• Model: https://huggingface.co/ornith-ai/Ornith-1.5-9B
• Files: https://huggingface.co/ornith-ai/Ornith-1.5-9B/tree/main

Get the model

Download its files with Hugging Face's command line:

pip install -U huggingface_hub
hf download ornith-ai/Ornith-1.5-9B

Then follow the usage in its model card. Nothing in this kit runs by itself.