Ornith 1.5 122B DFlash2 needs about 1.3 GB of memory at Q4_K_M (a 0.6 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 29 quants, from 0.1 GB (Q1_0) to 3.7 GB (F32). Longer context needs more: at 32K tokens the KV cache alone adds about 0.8 GB.
Ornith 1.5 122B DFlash2 – a block‑diffusion drafter for speculative decoding
Ornith 1.5 122B DFlash2 is a 0.9B drafter model that predicts whole blocks of tokens for a target Ornith‑1.5‑122B model. It is not a standalone language model and must be run with a target model via llama.cpp. The model supports a 262144‑token context.
Good at
- Speculative decoding draft
- Block‑diffusion token prediction
- Lightweight selector for coherent paths
Limits
- Requires target Ornith‑1.5‑122B GGUF
- No vision support
- Needs llama.cpp with DFlash 2 support
Best for
- Speeding up Ornith‑1.5‑122B inference
- Reducing token verification steps
- Optimizing speculative decoding pipelines
Who it’s for. Developers and researchers building speculative decoding systems that use Ornith‑1.5‑122B as the target model. The drafter is useful for those who need a lightweight draft model to reduce verification overhead.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q1_0 | 0.1 GB | 0.8 GB | Fits | Get file → | |
| TQ1_0 | 0.2 GB | 0.9 GB | Fits | Get file → | |
| TQ2_0 | 0.3 GB | 1.0 GB | Fits | Get file → | |
| Q2_0 | 0.3 GB | 1.0 GB | Fits | Get file → | |
| Q2_K | 0.3 GB | 1.0 GB | Fits | Get file → | |
| Q3_K_S | 0.4 GB | 1.1 GB | Fits | Get file → | |
| IQ3_S | 0.4 GB | 1.1 GB | Fits | Get file → | |
| IQ3_M | 0.4 GB | 1.1 GB | Fits | Get file → | |
| Q3_K | 0.4 GB | 1.1 GB | Fits | Get file → | |
| Q3_K_M | 0.4 GB | 1.1 GB | Fits | Get file → | |
| Q3_K_L | 0.5 GB | 1.2 GB | Fits | Get file → | |
| IQ4_XS | 0.5 GB | 1.2 GB | Fits | Get file → | |
| Q4_0 | 0.5 GB | 1.2 GB | Fits | Get file → | |
| Q4_K_S | 0.5 GB | 1.2 GB | Fits | Get file → | |
| IQ4_NL | 0.5 GB | 1.2 GB | Fits | Get file → | |
| Q4_K | 0.6 GB | 1.3 GB | Fits | Get file → | |
| Q4_K_M | 0.6 GB | 1.3 GB | Fits | Get file → | |
| Q4_1 | 0.6 GB | 1.3 GB | Fits | Get file → | |
| Q5_0 | 0.7 GB | 1.4 GB | Fits | Get file → | |
| Q5_K_S | 0.7 GB | 1.4 GB | Fits | Get file → | |
| Q5_K | 0.7 GB | 1.4 GB | Fits | Get file → | |
| Q5_K_M | 0.7 GB | 1.4 GB | Fits | Get file → | |
| Q5_1 | 0.7 GB | 1.4 GB | Fits | Get file → | |
| Q6_K | 0.8 GB | 1.5 GB | Fits | Get file → | |
| MXFP4_MOE | 1.0 GB | 1.7 GB | Fits | Get file → | |
| Q8_0 | 1.0 GB | 1.7 GB | Fits | Get file → | |
| BF16 | 1.9 GB | 2.6 GB | Fits | Get file → | |
| F16 | 1.9 GB | 2.6 GB | Fits | Get file → | |
| F32 | 3.7 GB | 4.4 GB | Fits | Get file → |
For 8 GB VRAM, use Q4_K_M (file size 0.55 GB, estimated memory 1.3 GB at 8 k context). For 4 GB VRAM, Q3_K_S (0.41 GB, 1.1 GB). For 2 GB VRAM, Q2_0 (0.27 GB, 1 GB).
What GPU do you need for Ornith 1.5 122B DFlash2?
These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.
- 4 GB VRAM: the best quant that fits is BF16 (1.9 GB file, about 2.6 GB in use).
- 6 GB VRAM: the best quant that fits is F32 (3.7 GB file, about 4.4 GB in use).
Which Ornith 1.5 122B DFlash2 quant should you choose?
Q4_K_M is the usual starting point for Ornith 1.5 122B DFlash2: it balances file size and quality well. Here it is a 0.6 GB file.
The Q8_0 file is 1.0 GB, about 1.8 times the size of Q4_K_M. It keeps the weights closer to the original, which only matters if you have the memory to spare.
If memory is tight, the smallest file is Q1_0 at 0.1 GB, with the quality trade-offs that come with fewer bits. Lower bit counts mean smaller files and more quality loss.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about Ornith 1.5 122B DFlash2
How much VRAM does Ornith 1.5 122B DFlash2 need?
About 1.3 GB at Q4_K_M with an 8K context window. That is the 0.6 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.
Can I run Ornith 1.5 122B DFlash2 on an 8 GB GPU?
Yes, with F32: the 3.7 GB file needs about 4.4 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of Ornith 1.5 122B DFlash2 GGUF?
It depends on the quant: Q1_0 is 0.1 GB, TQ1_0 is 0.2 GB, TQ2_0 is 0.3 GB, Q2_0 is 0.3 GB, Q2_K is 0.3 GB, Q3_K_S is 0.4 GB and 23 more.
What context length does Ornith 1.5 122B DFlash2 support?
The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 0.2 GB at 8K and 0.8 GB at 32K.
Where can I download Ornith 1.5 122B DFlash2 GGUF?
The files are on Hugging Face in the kaivoss/Ornith-1.5-122B-DFlash2-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does Ornith 1.5 122B DFlash2 use?
Hugging Face lists the license as apache-2.0. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.
Can I run Ornith 1.5 122B DFlash2 on 8 GB VRAM?
Yes, using Q4_K_M or Q3_K_S quants keeps memory below 8 GB.
Does this model support vision tasks?
No, it is a text‑only drafter and does not provide vision capabilities.
What is the maximum context length?
The model supports 262144 tokens of context.