Flash Update: DeepSeek has quietly opened internal testing for an intermediate V4.1 Flash release.
No application or form required — just change the model name and you can use it.
Key Information
| Item | Details |
|---|---|
| Model name | deepseek-v4.1-flash-expires-on-0910 |
| base_url | Unchanged — use the original one |
| Pricing | Same as deepseek-v4-flash |
| Concurrency limit | 20 concurrent requests per account (Note: the official release allows 2,500; the internal test version is heavily rate-limited) |
| Expiration | As the name suggests, expires on September 10 — try it while you can |
How do you get started?
Just change the model name used in your code to the one above:
cURL example
curl https://api.deepseek.com/v1/chat/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer $DEEPSEEK_API_KEY" \
-d '{
"model": "deepseek-v4.1-flash-expires-on-0910",
"messages": [{"role": "user", "content": "你好"}]
}'
Python (OpenAI SDK)
from openai import OpenAI
client = OpenAI(
api_key="你的API_KEY",
base_url="https://api.deepseek.com/v1"
)
response = client.chat.completions.create(
model="deepseek-v4.1-flash-expires-on-0910", # Just change this line
messages=[{"role": "user", "content": "介绍一下新模型"}]
)
What’s different about the new model?
According to the official description, it features a new model architecture and native multimodal support, making it more capable, faster, and cheaper. We’ll have to wait for everyone’s hands-on testing and feedback for the specific results.
Notes
- This is a time-limited test endpoint that expires on September 10, so don’t hard-code it into production environments
- With a 20-request concurrency limit, it’s fine for testing, but don’t use it for large-scale batch jobs
- The official documentation hasn’t been updated with a formal release record yet — it’s essentially a „sneak preview within the group“
Looking forward to everyone’s in-depth reviews — let’s see just how fast this version of Flash really is