Instructions to use incoai/Ornith-1.5-122B-DFlash2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use incoai/Ornith-1.5-122B-DFlash2 with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="incoai/Ornith-1.5-122B-DFlash2")

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModel

tokenizer = AutoTokenizer.from_pretrained("incoai/Ornith-1.5-122B-DFlash2")
model = AutoModel.from_pretrained("incoai/Ornith-1.5-122B-DFlash2", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use incoai/Ornith-1.5-122B-DFlash2 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "incoai/Ornith-1.5-122B-DFlash2"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "incoai/Ornith-1.5-122B-DFlash2",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/incoai/Ornith-1.5-122B-DFlash2

SGLang

How to use incoai/Ornith-1.5-122B-DFlash2 with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "incoai/Ornith-1.5-122B-DFlash2" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "incoai/Ornith-1.5-122B-DFlash2",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "incoai/Ornith-1.5-122B-DFlash2" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "incoai/Ornith-1.5-122B-DFlash2",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use incoai/Ornith-1.5-122B-DFlash2 with Docker Model Runner:
```
docker model run hf.co/incoai/Ornith-1.5-122B-DFlash2
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

Ornith-1.5-122B-DFlash2

Blog | GitHub

This repository contains the DFlash 2 draft model for ornith-ai/Ornith-1.5-122B. It is not a standalone language model: it runs inside a speculative decoding server and drafts tokens for the target model to verify.

DFlash 2 is a block-diffusion drafter for speculative decoding. It predicts a whole block of tokens in a single pass and keeps the top candidates at every position. A lightweight selector then traces one coherent path through them. Two-tap dynamic convolutions in the backbone keep the draft from decaying toward the end of the block. Decoding is lossless: greedy output matches the target model exactly, and sampling preserves its distribution.

DFlash 2: parallel block drafting with a candidate path selector

Quick Start

Serve with SGLang 0.5.21:

pip install "sglang[all]==0.5.21"

sglang serve \
  --model-path ornith-ai/Ornith-1.5-122B \
  --tp 2 \
  --enable-flashinfer-allreduce-fusion \
  --attention-backend trtllm_mha \
  --mamba-radix-cache-strategy extra_buffer \
  --mem-fraction-static 0.8 \
  --speculative-algorithm DFLASH \
  --speculative-draft-model-path incoai/Ornith-1.5-122B-DFlash2 \
  --speculative-num-draft-tokens 16

See the blog post for more details.

Evaluation

Runtime: SGLang 0.5.21 on two NVIDIA GB300 GPUs (TP2), with TensorRT-LLM for target attention and FlashAttention 4 for DFlash 2 draft attention
Speculation block size: 16 (15 draft tokens per verification step)
Sampling: Ornith's default parameters (temperature 0.6, top-p 0.95, top-k 20), with thinking enabled
Maximum new tokens: 4096
Samples: 128 per task
Prompts: benchmark formatting from z-lab/dflash; MT-Bench uses the first turn only

We compare autoregressive decoding and DFlash 2, which proposes fifteen draft tokens per verification step.

Acceptance Length

Acceptance length is the per-request mean of completion tokens divided by verification steps at concurrency 1. Higher is better.

Task	DFlash 2
GSM8K	8.33
MATH-500	9.20
HumanEval	8.64
MBPP	6.37
MT-Bench	5.41

Throughput

Throughput is total output tokens divided by end-to-end wall time. Each cell shows output tok/s (speedup vs. autoregressive).

Concurrency 1

Task	Autoregressive	DFlash 2
GSM8K	257.9	585.5 (2.27×)
MATH-500	265.8	705.5 (2.65×)
HumanEval	265.2	672.1 (2.53×)
MBPP	262.4	519.4 (1.98×)
MT-Bench	265.0	355.0 (1.34×)

Concurrency 8

Task	Autoregressive	DFlash 2
GSM8K	994.8	1,866.8 (1.88×)
MATH-500	1,020.0	2,378.4 (2.33×)
HumanEval	1,076.3	2,295.0 (2.13×)
MBPP	984.1	1,562.3 (1.59×)
MT-Bench	1,001.9	1,173.5 (1.17×)

Concurrency 32

Task	Autoregressive	DFlash 2
GSM8K	1,358.5	4,043.1 (2.98×)
MATH-500	1,845.4	4,440.1 (2.41×)
HumanEval	1,930.6	4,811.0 (2.49×)
MBPP	1,570.7	3,124.7 (1.99×)
MT-Bench	1,691.6	2,287.1 (1.35×)

Citation

If you find DFlash 2 useful, please cite:

@misc{inco2026dflash2,
  title  = {{DFlash 2: Keep Drafting Parallel}},
  author = {{Inco AI}},
  year   = {2026},
  month  = {August},
  url    = {https://inco.ai/blog/dflash2/}
}

Please also cite the original DFlash paper:

@inproceedings{chen2026dflash,
  title     = {{DFlash: Block Diffusion for Flash Speculative Decoding}},
  author    = {Chen, Jian and Liang, Yesheng and Liu, Zhijian},
  booktitle = {International Conference on Machine Learning (ICML)},
  year      = {2026}
}