DeepSeek-V4-Flash-0731

技术报告 👁️

简介

DeepSeek-V4-Flash-0731 是 DeepSeek-V4-Flash 的正式发布版本，取代预览版，并显著增强了 Agent 能力。它与 DeepSeek-V4-Flash-DSpark 采用相同的模型结构，即附带一个推测解码模块。

尽管激活参数量小得多，DeepSeek-V4-Flash-0731 在下列基准测试中的表现优于 DeepSeek-V4-Pro（Preview），并且整体上可与当前最强的专有模型竞争。

注释：

1. 对于上述公开基准中的 Code Agent 任务，DeepSeek-V4-Flash-0731 使用 DeepSeek Harness（即将发布）的最小模式作为 Agent 框架进行评测，采用 max 推理强度档位，并使用 temperature = 1.0, top_p = 0.95。
2. † DSBench-FullStack 是内部全栈开发测试集；DSBench-Hard 是内部高难度编码 Agent 问题测试集。

聊天模板

本次发布不包含 Jinja 格式的聊天模板。我们提供了专用的 encoding 文件夹，其中包含 Python 脚本和测试用例，演示如何将 OpenAI 兼容格式的消息编码为模型输入字符串，以及如何解析模型的文本输出。完整文档请参阅 encoding 文件夹。

reasoning_effort 参数现支持三个级别：low、high 和 max，用于控制模型在回答前投入多少推理过程。

简要示例：

from encoding_dsv4 import encode_messages, parse_message_from_completion_text

messages = [ {"role": "user", "content": "hello"}, {"role": "assistant", "content": "Hello! I am DeepSeek.", "reasoning_content": "thinking..."}, {"role": "user", "content": "1+1=?"} ]

prompt = encode_messages(messages, thinking_mode="thinking", reasoning_effort="max")

import transformers tokenizer = transformers.AutoTokenizer.from_pretrained("deepseek-ai/DeepSeek-V4-Flash-0731") tokens = tokenizer.encode(prompt)

如何通过 vLLM 运行

DSpark 推测解码可通过单个标志启用——在 vLLM 启动命令中添加带有 method: dspark 的 --speculative-config：

--speculative-config '{"method":"dspark","num_speculative_tokens":7,"draft_sample_method":"greedy"}'

例如，以下命令会在单个 4×GB300 节点上通过 vLLM 提供模型服务。详细说明及其他硬件配置请参阅 vLLM 配方。

vllm serve deepseek-ai/DeepSeek-V4-Flash-0731
--trust-remote-code --kv-cache-dtype fp8 --block-size 256
--data-parallel-size 4 --enable-expert-parallel
--moe-backend deep_gemm_mega_moe
--attention-config '{"use_fp4_indexer_cache": true}'
--speculative-config '{"method":"dspark","num_speculative_tokens":7,"draft_sample_method":"greedy"}'

如何通过 SGLang 运行

使用 --speculative-algorithm DSPARK 启用 DSpark，且不要单独设置 --speculative-draft-model-path，因为目标模型和草稿模型的权重来自同一个 checkpoint。详细说明、基准测试和其他硬件配置请参阅 SGLang cookbook。

sglang serve
--trust-remote-code
--model-path deepseek-ai/DeepSeek-V4-Flash-0731
--tp 4
--moe-runner-backend flashinfer_mxfp4
--speculative-algorithm DSPARK
--mem-fraction-static 0.90
--chunked-prefill-size 4096
--swa-full-tokens-ratio 0.1 \

如何在本地运行

有关在本地运行 DeepSeek-V4 的详细说明，包括模型权重转换和交互式聊天演示，请参阅 inference 文件夹。

对于本地部署，我们建议将采样参数设置为 temperature = 1.0；对于 Agent 场景使用 top_p = 0.95，其他场景则使用 top_p = 1.0。对于 high 和 max 推理强度档位，建议将最大输出长度设置为 384K tokens。

许可证

此仓库及模型权重采用 MIT License 许可。

引用

@misc{deepseekai2026deepseekv4, title={DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence}, author={DeepSeek-AI}, year={2026}, }

联系方式

如有任何问题，请提交 issue，或通过 service@deepseek.com 联系我们。

DeepSeek-V4-Flash-0731

Technical Report 👁️

Introduction

DeepSeek-V4-Flash-0731 is the official release of DeepSeek-V4-Flash , superseding the preview version, with substantially enhanced agentic capabilities. It has the same model structure as DeepSeek-V4-Flash-DSpark, i.e. it comes with a speculative decoding module attached.

DeepSeek-V4-Flash-0731 outperforms DeepSeek-V4-Pro (Preview) on benchmarks listed below despite its far smaller activated parameter count, and is broadly competitive with the strongest proprietary models available.

Notes:

1. For the Code Agent tasks among the public benchmarks above, DeepSeek-V4-Flash-0731 is evaluated with the minimal mode of DeepSeek Harness (to be released) as the agent framework, using the max reasoning effort level with temperature = 1.0, top_p = 0.95.
2. † DSBench-FullStack is an internal full-stack development test set; DSBench-Hard is an internal test set of difficult coding-agent problems.

Chat Template

This release does not include a Jinja-format chat template. Instead, we provide a dedicated encoding folder with Python scripts and test cases demonstrating how to encode messages in OpenAI-compatible format into input strings for the model, and how to parse the model's text output. Please refer to the encoding folder for full documentation.

The reasoning_effort parameter now supports three levels — low, high, and max — which control how much deliberation the model spends before answering.

A brief example:

from encoding_dsv4 import encode_messages, parse_message_from_completion_text

messages = [
    {"role": "user", "content": "hello"},
    {"role": "assistant", "content": "Hello! I am DeepSeek.", "reasoning_content": "thinking..."},
    {"role": "user", "content": "1+1=?"}
]


prompt = encode_messages(messages, thinking_mode="thinking", reasoning_effort="max")


import transformers
tokenizer = transformers.AutoTokenizer.from_pretrained("deepseek-ai/DeepSeek-V4-Flash-0731")
tokens = tokenizer.encode(prompt)

How to Run with vLLM

DSpark speculative decoding is enabled with a single flag — add --speculative-config with method: dspark to your vLLM launch command:

--speculative-config '{"method":"dspark","num_speculative_tokens":7,"draft_sample_method":"greedy"}'

For example, the command below serves the model with vLLM on a single 4×GB300 node. See the vLLM recipe for detailed instructions and other hardware configurations.

vllm serve deepseek-ai/DeepSeek-V4-Flash-0731 \
  --trust-remote-code --kv-cache-dtype fp8 --block-size 256 \
  --data-parallel-size 4 --enable-expert-parallel \
  --moe-backend deep_gemm_mega_moe \
  --attention-config '{"use_fp4_indexer_cache": true}' \
  --speculative-config '{"method":"dspark","num_speculative_tokens":7,"draft_sample_method":"greedy"}'

How to Run with SGLang

Enable DSpark with --speculative-algorithm DSPARK and do not set a separate --speculative-draft-model-path as the target and draft weights therefore come from the same checkpoint. See the SGLang cookbook for detailed instructions, benchmarks and other hardwares configurations.

sglang serve \
  --trust-remote-code \
  --model-path deepseek-ai/DeepSeek-V4-Flash-0731 \
  --tp 4 \
  --moe-runner-backend flashinfer_mxfp4 \
  --speculative-algorithm DSPARK \
  --mem-fraction-static 0.90 \
  --chunked-prefill-size 4096 \
  --swa-full-tokens-ratio 0.1 \

How to Run Locally

Please refer to the inference folder for detailed instructions on running DeepSeek-V4 locally, including model weight conversion and interactive chat demos.

For local deployment, we recommend setting the sampling parameters to temperature = 1.0, with top_p = 0.95 for agentic scenarios and top_p = 1.0 otherwise. For the high and max reasoning effort levels, we recommend a maximum output length of 384K tokens.

License

This repository and the model weights are licensed under the MIT License.

Citation

@misc{deepseekai2026deepseekv4,
      title={DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence},
      author={DeepSeek-AI},
      year={2026},
}

Contact

If you have any questions, please raise an issue or contact us at service@deepseek.com.