On this page
You can run Ornith 1.5 9B on an Android phone with 12 GB of RAM or more, using Off Grid AI Mobile (OGAM). Search for Ornith, download ornith-ai’s 4-bit build (about 6.7 GB with its vision file), and you have a 9B reasoning model running on your phone with no internet and no account.
It’s free, and the model is MIT-licensed. On an 8 GB phone it won’t fit, and this guide says what to run instead.
Google Play · Download Off Grid AI
What Ornith 1.5 is
Ornith 1.5 comes from ornith-ai, a team that publishes on Hugging Face. The family has a 9B dense model, a 35B-A3B mixture of experts, a 122B and a 397B, all under the MIT licence. The 9B and 35B-A3B GGUF builds went up in August 2026, and the 9B GGUF has about 5.8 million downloads.
The model card says Ornith 1.0 was built on top of Qwen 3.5 and Gemma 4 with more training, and 1.5 adds a self-improvement loop where the model helps make its own training tasks. The 9B GGUF file itself is a Qwen 3.5 model in structure: its header names the qwen35 architecture, the same one as the Qwen 3.5 models in OGAM’s catalog.
ornith-ai reports these scores for the 9B against Qwen 3.5 9B. They’re the team’s own numbers, averaged over five runs.
| Benchmark | Ornith 1.5 9B | Qwen 3.5 9B |
|---|---|---|
| SWE-bench Verified | 70.6 | 53.2 |
| GPQA Diamond | 86.4 | 81.7 |
| HLE (no tools) | 20.2 | 14.7 |
It’s a reasoning model. Each answer starts with a thinking step, so expect a pause before the reply begins.
Which phones it fits
OGAM judges fit from your phone’s RAM. On Android, a download under 52% of RAM runs well, 52% to 62% is marked Tight, and anything bigger isn’t offered. The download is the model file plus its vision file, because the app fetches both.
| File | Download with vision file | 8 GB phone | 12 GB phone | 16 GB phone |
|---|---|---|---|---|
| Q4_K_M | 6.70 GB (5.78 + 0.92) | Not offered | Tight | Runs well |
| Q5_K_M | 7.56 GB (6.64 + 0.92) | Not offered | Not offered | Runs well |
Sizes come from ornith-ai’s GGUF repo. There’s no smaller file than Q4_K_M from ornith-ai, which is why 8 GB phones are out. To check your RAM, open Android Settings, then About phone.
Tight means it loads but leaves little room. Close other apps before a long chat on a 12 GB phone.
Set it up
- Install OGAM from Google Play.
- Tap Models in the bottom bar. You land on the Text tab.
- Type Ornith in Search models.
- Tap Ornith-1.5-9B-GGUF from ornith-ai. Don’t confuse it with Ornith-1.0-9B-GGUF, the older version, which shows in the same results.
- Under Sizes, take the Recommended file and download it. Use Wi-Fi, because it’s about 6.7 GB.
- When it finishes, tap Chat now.
Ornith isn’t in OGAM’s own catalog, so it comes from Hugging Face search like any other GGUF. We haven’t timed it on a phone, so we don’t quote a speed. A 9B dense model is slower than the 2B and 4B models most phones run.
Once it’s downloaded, turn on airplane mode and ask again. It still answers, because the model runs on your phone.
What to use it for
ornith-ai trained Ornith 1.5 for coding, reasoning and agent work, and its scores lead on coding. On a phone, that means:
- Read a stack trace you pasted and explain what broke.
- Explain a function from a code review you’re reading on the train.
- Work through a maths or logic problem step by step.
- Plan a task into ordered steps before you start it.
For quick replies and everyday writing, a smaller model answers faster. Ornith is the one to load when the question is hard.
On an 8 GB phone
Run Qwen 3.5 4B on the phone instead. It’s in OGAM’s catalog and shares Ornith’s base architecture.
If you have a computer, you can also run Ornith there, in Off Grid AI Desktop or Ollama, and chat with it from your phone over your own Wi-Fi. The phone-to-computer guide shows the Remote servers setup.
If something goes wrong
| Symptom | Check | Fix |
|---|---|---|
| No Ornith 1.5 result | Your phone has under 12 GB of RAM | Use Qwen 3.5 4B on the phone, or run Ornith on a computer |
| The app closes when the model loads | It’s Tight on 12 GB and other apps hold memory | Close other apps and load it again |
| A long pause before each reply | It’s a reasoning model and thinks first | Ask for a short answer, or use a smaller model for quick questions |
FAQ
Can Ornith 1.5 9B run on an Android phone? Yes, on a phone with 12 GB of RAM or more. Search for Ornith in OGAM and download ornith-ai’s own GGUF. It uses the same Qwen 3.5 architecture as the Qwen 3.5 models the app already runs.
How big is the Ornith 1.5 9B download? The 4-bit Q4_K_M file is 5.78 GB and the vision file is 0.92 GB, so about 6.7 GB in all. OGAM marks it Tight on a 12 GB phone and Runs well on a 16 GB phone.
Will Ornith 1.5 9B run on an 8 GB phone? No. Its smallest published file is too big for 8 GB of RAM, so the app doesn’t offer it. On an 8 GB phone, run Qwen 3.5 4B instead.
Is Ornith 1.5 free to use? Yes. ornith-ai releases it under the MIT licence, and OGAM is free with no account.
Is there a mobile version of Ornith 1.5? The model card mentions an Ornith-1.5-9B-Mobile variant, but as of 16 November 2026 ornith-ai hasn’t published it on Hugging Face. The regular 9B GGUF is what runs on phones today.
Related guides
- How to Run Qwen 3.5 on Your Android Phone in 2026 (Locally, No Cloud)
- How to Find AI Models Your Android Phone Can Actually Run in 2026 (No Failed Downloads)
- How to Fix a Local AI Model That Crashes on Android in 2026
- How to Use AI Models on Your Mac From Your Android Phone in 2026
Get Off Grid AI on Google Play and load Ornith 1.5 9B on your 12 GB phone.