Ornith 1.5 122B VRAM requirements
Ornith AI · 122B params (10B active) · MIT. Estimates, not vendor numbers.
Ornith 1.5 122B (122B parameters, 10B active per token) needs about 75 GB of VRAM or unified memory at Q4, 88 GB at Q5 and 131 GB at Q8 (weights plus about 1.5 GB headroom, our estimate). The smallest common setup that runs it is Mac 128GB (M4 Max).
| Quant | Download | Memory needed |
|---|---|---|
| Q4 | 73 GB | 75 GB |
| Q5 | 87 GB | 88 GB |
| Q8 | 130 GB | 131 GB |
Will it run on my machine?
| Setup | Runs? | Speed est. |
|---|---|---|
| 8GB GPU (RTX 4060) | No | n/a |
| 12GB GPU (RTX 3060 / 4070) | No | n/a |
| 16GB GPU (RTX 4060 Ti 16GB) | No | n/a |
| 24GB GPU (RTX 4090) | No | n/a |
| 32GB GPU (RTX 5090) | No | n/a |
| Mac 16GB (M4) | No | n/a |
| Mac 48GB (M4 Pro) | No | n/a |
| Mac 128GB (M4 Max) | Yes, Q5 | ~46 tok/s |
| PC, 32GB RAM, no GPU | No | n/a |
Best quant that fits each setup, with estimated generation speed from memory bandwidth. Real speed depends on the runtime, context length and settings.
Mixture of experts built on Alibaba's Qwen3.5 122B-A10B architecture, so about 10B active per token (Ornith's card doesn't state the active count), 262K context. Weights posted Oct 16, 2026. Ornith says it serves on a single 80 GB GPU at full precision; 4-bit builds need far less.
FAQ
How much VRAM does Ornith 1.5 122B need?
About 75 GB at Q4, 88 GB at Q5 and 131 GB at Q8, including about 1.5 GB headroom. These are estimates from the parameter count on the official model card.
Can Ornith 1.5 122B run on 8GB of VRAM?
Not fully on the GPU. At Q4 it needs about 75 GB; partial CPU offload works but is much slower.
Can Ornith 1.5 122B run on a Mac?
Yes. Mac 128GB (M4 Max) can run it.
Which quant of Ornith 1.5 122B should I use?
Use the highest quant that fits with room to spare: Q8 is near full quality, Q5 is a good balance, and Q4 is the smallest that usually keeps quality acceptable.
Why is Ornith 1.5 122B fast for its size?
It is a mixture of experts: 122B parameters in total but only 10B are used per token, so it reads less memory per token. You still need memory for all 122B.
Similar-size models
Qwen3.6 35B A3B (35B) · Recursive Flint 40B A2.5B (39.8B) · Llama 3.3 70B Instruct (70B) · gpt-oss-120b (117B) · Mistral Small 4 (119B) · Qwen3.8 Flash Next (180B) · Muse Spark Open (816B)