Total runs:
2.4M
Run Growth:
284.6K
Growth Rate:
11.91%
Sponsored by
YouArt
deepseek-ai / DeepSeek-V4-Flash-0731
Model's Last Updated: August 01 2026
text-generation
Introduction of DeepSeek-V4-Flash-0731
Model Details of DeepSeek-V4-Flash-0731
DeepSeek-V4-Flash-0731
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.
| Benchmark | DeepSeek-V4-Flash-0731 | DeepSeek-V4-Flash (Preview) | DeepSeek-V4-Pro (Preview) | GLM-5.2 | Opus-4.8 |
|---|---|---|---|---|---|
| Terminal Bench 2.1 | 82.7 | 61.8 | 72.1 | 81.0 | 85.0 |
| NL2Repo | 54.2 | 39.4 | 38.5 | 48.9 | 69.7 |
| Cybergym | 76.7 | 38.7 | 52.7 | - | 83.1 |
| DeepSWE | 54.4 | 7.3 | 12.8 | 46.2 | 58.0 |
| Toolathlon-Verified | 70.3 | 49.7 | 55.9 | 59.9 | 76.2 |
| Agents' Last Exam | 25.2 | 15.8 | 16.5 | 23.8 | 25.7 |
| AutomationBench Public | 25.1 | 10.8 | 12.8 | 12.9 | 27.2 |
| DSBench-FullStack † | 68.7 | 37.0 | 41.8 | 61.8 | 71.6 |
| DSBench-Hard † | 59.6 | 25.8 | 31.1 | 54.5 | 71.7 |
Notes:
-
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
maxreasoning effort level withtemperature = 1.0, top_p = 0.95. - † 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=?"}
]
# messages -> string
prompt = encode_messages(messages, thinking_mode="thinking", reasoning_effort="max")
# string -> tokens
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 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 [email protected] .
Runs of deepseek-ai DeepSeek-V4-Flash-0731 on huggingface.co
4.4M
Total runs
-23.0K
24-hour runs
-51.7K
3-day runs
-112.5K
7-day runs
2.9M
30-day runs
More Information About DeepSeek-V4-Flash-0731 huggingface.co Model
Url of DeepSeek-V4-Flash-0731
DeepSeek-V4-Flash-0731 huggingface.co Url
Provider of DeepSeek-V4-Flash-0731 huggingface.co
Other API from deepseek-ai
Updated:November 04 2025
Total runs:
1.8M
Run Growth:
679.6K
Growth Rate:
37.66%
Total runs:
1.8M
Run Growth:
-425.6K
Growth Rate:
-23.81%
Total runs:
959.0K
Run Growth:
-388.4K
Growth Rate:
-40.50%
Total runs:
933.6K
Run Growth:
-516.1K
Growth Rate:
-55.28%
Total runs:
920.4K
Run Growth:
-203.5K
Growth Rate:
-22.11%
Total runs:
802.4K
Run Growth:
284.4K
Growth Rate:
35.44%
Total runs:
783.1K
Run Growth:
175.6K
Growth Rate:
22.57%
Total runs:
736.0K
Run Growth:
32.2K
Growth Rate:
4.38%
Total runs:
682.1K
Run Growth:
-7.6M
Growth Rate:
-1119.23%
Total runs:
619.4K
Run Growth:
-766.2K
Growth Rate:
-123.69%
Total runs:
592.0K
Run Growth:
-185.9K
Growth Rate:
-31.40%
Total runs:
540.6K
Run Growth:
168.9K
Growth Rate:
31.24%
Total runs:
507.6K
Run Growth:
-147.4K
Growth Rate:
-29.05%
Total runs:
484.4K
Run Growth:
466.5K
Growth Rate:
96.31%
Total runs:
401.0K
Run Growth:
-231.5K
Growth Rate:
-57.74%
Total runs:
397.8K
Run Growth:
-127.8K
Growth Rate:
-32.25%
Total runs:
361.0K
Run Growth:
99.9K
Growth Rate:
27.67%
Total runs:
344.4K
Run Growth:
-486.0K
Growth Rate:
-141.11%
Total runs:
343.7K
Run Growth:
-104.1K
Growth Rate:
-30.30%
Total runs:
312.5K
Run Growth:
-52.1K
Growth Rate:
-16.69%
Total runs:
242.6K
Run Growth:
49.6K
Growth Rate:
20.46%
Total runs:
242.1K
Run Growth:
10.2K
Growth Rate:
4.23%
Total runs:
176.5K
Run Growth:
176.5K
Growth Rate:
100.00%
Total runs:
154.0K
Run Growth:
-1.0M
Growth Rate:
-661.31%
Total runs:
140.6K
Run Growth:
140.6K
Growth Rate:
100.00%
Total runs:
77.0K
Run Growth:
-140.9K
Growth Rate:
-182.93%
Total runs:
73.5K
Run Growth:
71.8K
Growth Rate:
97.55%
Total runs:
67.1K
Run Growth:
-508
Growth Rate:
-0.76%
Total runs:
59.7K
Run Growth:
-110.5K
Growth Rate:
-185.07%
Total runs:
43.5K
Run Growth:
27.4K
Growth Rate:
62.85%
Total runs:
36.6K
Run Growth:
-69.0K
Growth Rate:
-191.56%
Total runs:
30.5K
Run Growth:
-3.4K
Growth Rate:
-11.11%
Total runs:
29.9K
Run Growth:
20.0K
Growth Rate:
66.89%
Total runs:
28.5K
Run Growth:
12.3K
Growth Rate:
43.17%
Total runs:
28.0K
Run Growth:
8.8K
Growth Rate:
31.39%
Total runs:
23.8K
Run Growth:
-12.4K
Growth Rate:
-52.00%
Total runs:
20.9K
Run Growth:
-27.5K
Growth Rate:
-131.25%
Total runs:
20.5K
Run Growth:
9.0K
Growth Rate:
43.78%
Total runs:
17.7K
Run Growth:
104
Growth Rate:
0.59%
Total runs:
13.7K
Run Growth:
7.8K
Growth Rate:
56.53%
Total runs:
10.7K
Run Growth:
-15.9K
Growth Rate:
-148.02%
Total runs:
10.4K
Run Growth:
2.6K
Growth Rate:
24.49%
Total runs:
10.3K
Run Growth:
3.5K
Growth Rate:
34.07%
Total runs:
10.1K
Run Growth:
-2.8K
Growth Rate:
-28.20%
Total runs:
10.0K
Run Growth:
820
Growth Rate:
8.23%
Total runs:
9.7K
Run Growth:
4.0K
Growth Rate:
41.17%
Total runs:
9.6K
Run Growth:
-7.1K
Growth Rate:
-72.03%
Total runs:
9.6K
Run Growth:
-4.5K
Growth Rate:
-45.68%
Total runs:
9.3K
Run Growth:
609
Growth Rate:
6.58%
Total runs:
8.5K
Run Growth:
2.1K
Growth Rate:
24.50%
Total runs:
8.3K
Run Growth:
5.7K
Growth Rate:
68.51%
Total runs:
8.0K
Run Growth:
-741
Growth Rate:
-9.23%
Total runs:
7.5K
Run Growth:
2.6K
Growth Rate:
35.08%
Total runs:
7.2K
Run Growth:
-663
Growth Rate:
-9.20%
Total runs:
7.1K
Run Growth:
1.5K
Growth Rate:
21.53%
Total runs:
7.1K
Run Growth:
-26.0K
Growth Rate:
-368.07%
Total runs:
6.4K
Run Growth:
4.3K
Growth Rate:
66.99%
Total runs:
5.7K
Run Growth:
2.3K
Growth Rate:
41.58%
Total runs:
5.2K
Run Growth:
3.6K
Growth Rate:
70.08%
Total runs:
5.0K
Run Growth:
2.3K
Growth Rate:
45.83%
Total runs:
3.8K
Run Growth:
1.0K
Growth Rate:
26.63%
Total runs:
3.6K
Run Growth:
-55
Growth Rate:
-1.51%
Total runs:
3.5K
Run Growth:
-19.3K
Growth Rate:
-554.12%
Total runs:
3.4K
Run Growth:
-322
Growth Rate:
-9.55%
Total runs:
2.8K
Run Growth:
-4
Growth Rate:
-0.15%
Total runs:
2.0K
Run Growth:
649
Growth Rate:
32.68%
Total runs:
1.9K
Run Growth:
-848
Growth Rate:
-45.32%
Total runs:
1.8K
Run Growth:
713
Growth Rate:
39.03%
Total runs:
1.8K
Run Growth:
354
Growth Rate:
19.97%
Total runs:
1.5K
Run Growth:
226
Growth Rate:
14.93%
Total runs:
965
Run Growth:
116
Growth Rate:
12.02%
Total runs:
832
Run Growth:
-1.2K
Growth Rate:
-126.21%
Total runs:
805
Run Growth:
-233
Growth Rate:
-28.94%
Total runs:
770
Run Growth:
61
Growth Rate:
7.92%
Total runs:
701
Run Growth:
13
Growth Rate:
1.85%
Total runs:
689
Run Growth:
-38
Growth Rate:
-5.52%
Total runs:
675
Run Growth:
113
Growth Rate:
16.74%
Total runs:
652
Run Growth:
233
Growth Rate:
35.74%
Total runs:
619
Run Growth:
104
Growth Rate:
16.80%
Total runs:
617
Run Growth:
232
Growth Rate:
37.60%
Total runs:
561
Run Growth:
100
Growth Rate:
17.83%
Total runs:
453
Run Growth:
88
Growth Rate:
19.43%
Total runs:
415
Run Growth:
95
Growth Rate:
22.89%
Total runs:
375
Run Growth:
102
Growth Rate:
27.20%
Total runs:
370
Run Growth:
-379
Growth Rate:
-102.43%
Total runs:
329
Run Growth:
184
Growth Rate:
56.97%
Total runs:
326
Run Growth:
177
Growth Rate:
55.66%
Total runs:
323
Run Growth:
167
Growth Rate:
53.02%
Total runs:
323
Run Growth:
134
Growth Rate:
42.14%
Total runs:
316
Run Growth:
158
Growth Rate:
51.30%
Total runs:
315
Run Growth:
165
Growth Rate:
53.75%
Total runs:
315
Run Growth:
162
Growth Rate:
52.77%
Total runs:
309
Run Growth:
163
Growth Rate:
53.09%
Total runs:
305
Run Growth:
163
Growth Rate:
53.80%
Total runs:
303
Run Growth:
153
Growth Rate:
50.83%
Total runs:
303
Run Growth:
154
Growth Rate:
50.99%
Total runs:
302
Run Growth:
163
Growth Rate:
54.33%
Total runs:
301
Run Growth:
164
Growth Rate:
54.85%