DeepSeek-V4-Pro-0813 Publish
TL;DR Highlight
Change base_url and call DeepSeek with OpenAI or Anthropic SDKs.
Who Should Read
A backend developer who wants to switch the model behind the OpenAI SDK or agent tools like Claude Code to DeepSeek
Core Mechanics
- OpenAI- and Anthropic-compatible API
- Change base_url and you can reach DeepSeek with the OpenAI or Anthropic SDK.
- The OpenAI format uses https://api.deepseek.com and the Anthropic format uses https://api.deepseek.com/anthropic.
- Model names: deepseek-flash and deepseek-v4-pro
- deepseek-flash is now served by DeepSeek-V4.1-Flash-1017, and the calling method is unchanged.
- The legacy names deepseek-v4-flash and deepseek-v4-flash-vision-exp are still accepted, but those models are retired.
- Legacy names are billed at Flash price
- Requests using the old names are served by DeepSeek-V4.1-Flash-1017.
- They are billed at the Flash price.
- thinking and reasoning_effort parameters
- The examples set thinking to {"type":"enabled"} and reasoning_effort to "high".
- In the Python SDK, thinking is passed through extra_body.
- Works as a backend for agent tools
Community Reactions
- Naming confusion
- One commenter mocked the versioned, suffixed model naming and wished it were simply called 4.1.
- Doubts about benchmark source
- A commenter said the shared benchmarks reportedly came from a DeepSeek WeChat group, so take them with a grain of salt.
- Pricing worries
- A commenter wondered what pricing will look like after the planned "significant increase".
- No vision support
- A commenter was disappointed that vision appears to be missing.
How to Apply
- Point your existing OpenAI SDK call at DeepSeek
- Set base_url to https://api.deepseek.com and model to deepseek-flash in your current code, then make one call.
- Provide the API key via the DEEPSEEK_API_KEY environment variable.
- Replace legacy model names in your code
- Search for deepseek-v4-flash and deepseek-v4-flash-vision-exp and change them to deepseek-flash.
- Since the vision model is retired, check whether those calls actually depend on vision.
- Compare thinking on vs off
- Run the same prompt while varying the thinking and reasoning_effort settings, and compare quality and token usage.
- Use the results to pick a default per service.
Try It
# Run in terminal (pip install openai first, set DEEPSEEK_API_KEY env var)
import os
from openai import OpenAI
client = OpenAI(
api_key=os.environ.get('DEEPSEEK_API_KEY'),
base_url="https://api.deepseek.com",
)
response = client.chat.completions.create(
model="deepseek-flash",
messages=[
{"role": "system", "content": "You are a helpful assistant"},
{"role": "user", "content": "{{question}}"},
],
stream=False,
reasoning_effort="high",
extra_body={"thinking": {"type": "enabled"}},
)
print(response.choices[0].message.content)
# Next: set stream=True to try streaming responsesTerminology
Related Papers
Mercury 3: Diffusion at scale
Scaling diffusion LLM 40x gives Mercury 3: 4–8x faster than speed-tier models, similar quality
Timestral: Mistral's time-series foundation models
A 209M time-series model ranks first on three benchmarks and runs on CPU.
DeepSeek V4.1 Flash 1017 (Max)
DeepSeek V4.1 Flash 1017 Max scores 45 on the Intelligence Index at 211 tok/s and $0.30 input.
If you don't run your own LLM proxy, how do you know whats going on?
Running your own LLM proxy shows which keys, models, and cache misses drive your usage.
DeepSeek: DeepSeek v4.1 Flash 1017
DeepSeek V4.1 Flash 1017: $0.05/M input, 1M context, built for agents
DeepSeek-v4.1-Flash-1017
DeepSeek V4.1 Flash minor update lifts coding and agent benchmarks while cutting run time and output length
Show HN: Llmspeed – feel how fast a local LLM runs before you buy the hardware