nvidia/Kimi-K2.7-Code-DFlash

Model Overview

Description:

The NVIDIA Kimi-K2.7-Code DFlash model is the DFlash draft head of Moonshot AI's Kimi-K2.7-Code model, which is a coding-focused agentic language model that uses an optimized transformer architecture. For more information, please check here. The NVIDIA Kimi-K2.7-Code DFlash model incorporates DFlash speculative decoding with Model Optimizer.

This model is ready for commercial or non-commercial use.

License/Terms of Use:

Governing Terms: Use of this model is governed by the NVIDIA Open Model License.

ADDITIONAL INFORMATION : Modified MIT License. Kimi-K2.7-Code .

Deployment Geography:

Global

Use Case:

Developers designing AI Agent systems, coding assistants, software engineering copilots, and other AI-powered applications requiring long-horizon code generation and reasoning. Also suitable for typical instruction-following tasks where latency-optimized inference via speculative decoding is desirable.

Release Date:

Hugging Face 07/09/2026 via https://huggingface.co/nvidia/Kimi-K2.7-Code-DFlash

Reference(s):

• Kimi-K2.7-Code release notes

Model Architecture:

Architecture Type: Transformers

Network Architecture: DeepSeek V3

Number of Model Parameters: 1T in total and 32B activated

Input:

Input Type(s): Text, Image, Video

Input Format(s): String, Binary(Base64 encoded), Binary(Base64 encoded)

Input Parameters: One-Dimensional (1D), Two-Dimensional (2D), Three-Dimensional (3D)

Other Properties Related to Input: Context length: 256K

Output:

Output Type(s): Text

Output Format: String

Output Parameters: One Dimensional(1D): Sequences

Other Properties Related to Output: Outputs may include natural-language responses, code, structured JSON, tool-call requests, agent coordination instructions, and generated artifacts depending on serving configuration and application-level tooling.

Our AI models are designed and/or optimized to run on NVIDIA GPU-accelerated systems. By leveraging NVIDIA's hardware (e.g. GPU cores) and software frameworks (e.g., CUDA libraries), the model achieves faster training and inference times compared to CPU-only solutions.

Software Integration:

Supported Runtime Engine(s):

• vLLM

Supported Hardware Microarchitecture Compatibility:

• NVIDIA Blackwell B200

Preferred Operating System(s):

• Linux

The integration of foundation and fine-tuned models into AI systems requires additional testing using use-case-specific data to ensure safe and effective deployment. Following the V-model methodology, iterative testing and validation at both unit and system levels are essential to mitigate risks, meet technical and functional requirements, and ensure compliance with safety and ethical standards before deployment.

Model Version(s):

The model is DFlash version and is trained with nvidia-modelopt v0.45.0

Training and Evaluation Datasets:

Training Dataset:

Link: Nemotron-Post-Training-Dataset-v2, only prompts from the datasets were used for data synthesis, (the original responses from GPT were not used), which is then used to train the DFlash modules.

Data Modality: Text, Image, Video

Image Training Data Size: None

Text Training Data Size: [1 Billion to 10 Trillion Tokens]

Video Training Data Size: None

Data Collection Method by dataset: Hybrid: Automated, Synthetic

Labeling Method by dataset: Hybrid: Automated, Synthetic

Properties: 112K multilingual text samples featuring prompts spanning math, code, STEM, and conversational topics. Each sample includes a synthetic response generated by the target model.

Evaluation Dataset:

Link: MTBench, for more details, see here; SPEED-Bench

Data Collection Method by dataset: Hybrid: manually-collected, Synthetic

Labeling Method by dataset:Hybrid: manually-labelled, Synthetic

Properties: MT-Bench contains 3,300 multi-turn dialogue sequences, each annotated with expert preference votes. SPEED-Bench is a unified, diverse benchmark for speculative decoding spanning coding, humanities, math, multilingual, QA, RAG, reasoning, roleplay, STEM, summarization, and writing domains.

Inference:

Acceleration Engine: vLLM

Test Hardware: NVIDIA Blackwell B200

DFlash Speculative Decoding

Synthesized data was obtained from Moonshot AI's Kimi-K2.7-Code model, which is then used to finetune the DFlash modules. This model is ready for inference with vLLM in DFlash speculative decoding mode. DFlash modules are used to predict candidate tokens beyond the next token. In the generation step, each forward DFlash module generates a distribution of tokens beyond the previous. The longest accepted candidate sequence is selected so that more than 1 token is returned in the generation step. The number of tokens generated in each step is called acceptance rate.

Usage

To serve the checkpoint with vLLM:

vllm serve \
--tensor-parallel-size 4 \
--trust-remote-code \
--speculative-config '{
"method": "dflash",
"model": "",
"num_speculative_tokens":8
}'

Alternatively, with the Python LLM API:

from vllm import LLM, SamplingParams

llm = LLM(
model="",
tensor_parallel_size=4,
trust_remote_code=True,
speculative_config={
"method": "dflash",
"model": "",
"num_speculative_tokens":8
},
)

Evaluation

Acceptance rate on SPEED-Bench (qualitative subset) with a draft block size of 8:

| Category | SPEED-Bench Acceptance Rate | |-----------------|:-----------------------:| | coding | 3.41 | | humanities | 2.76 | | math | 3.24 | | multilingual | 3.89 | | qa | 2.82 | | rag | 3.68 | | reasoning | 3.26 | | roleplay | 2.36 | | stem | 2.84 | | summarization | 3.48 | | writing | 2.66 | | Overall Average | 3.13 |...