DeepSeek V3.1 is the latest, state-of-the-art hybrid-inference AI model from DeepSeek, blending “Think” and “Non-Think” modes within a single architecture. It’s the newer version of the DeepSeek V3 model with effecient hybrid reasoning.

​
How to use DeepSeek V3.1

Get started with this model in 10 lines of code! The model ID is deepseek-ai/DeepSeek-V3.1 and the pricing is $0.60 for input tokens and $1.70 for output tokens. 
from together import Together

client = Together()
resp = client.chat.completions.create(
model="deepseek-ai/DeepSeek-V3.1",
messages=[{"role":"user","content":"What are some fun things to do in New York?"}],
stream=True,
)
for tok in resp:
print(tok.choices[0].delta.content, end="", flush=True)
Current Limitations. The following features are not yet supported, but will be added soon: Function calling and JSON mode.

​
Hybrid Thinking

Here’s how to enable thinking in DeepSeek V3. 
from together import Together
client = Together()

stream = client.chat.completions.create(
model="deepseek-ai/DeepSeek-V3.1",
messages=[
{"role": "user", "content": "What are some fun things to do in New York?"}
],
chat_template_kwargs={"thinking": True},
stream=True,
)

for chunk in stream:
delta = chunk.choices[0].delta

  # Show reasoning tokens if present
  if hasattr(delta, "reasoning") and delta.reasoning:
      print(delta.reasoning, end="", flush=True)

  # Show content tokens if present
  if hasattr(delta, "content") and delta.content:
      print(delta.content, end="", flush=True)

​
How is it different from DeepSeek V3?

DeepSeek V3.1 – the newer better version of DeepSeek V3 – has a few key differences:
  • Hybrid model w/ two main modes: Non-thinking and Thinking mode
  • Function calling only works in non-thinking mode
  • Agent capabilities: Built-in support for code agents and search agents
  • More efficient reasoning than DeepSeek-R1
  • Continued long-context pre-training