#31
DeepSeek V3.1
deepseek-chat

DeepSeek V3.1 is a hybrid inference model that serves as the predecessor to the experimental V3.2 line. It introduced the concept of unified weights for both thinking and non-thinking modes. In its standard configuration, it offers fast, cost-effective performance suitable for general-purpose chat and agentic tasks, with a 128k context window and strong general knowledge capabilities.

Performance Metrics

Helpfulness
Instruction Following
Comprehension
Empathy
Creative Writing
79.60
Helpfulness
83
Empathy
76
Instruction Following
84
Creative Writing
73
Comprehension
82
Speed
Avg 92 tok/s

Model Specifications

Release Date
August 21, 2025
Lab
DeepSeek
Type
Open Source
Context Size
128K
Max Output Tokens
8.2K
Cost per 1M tokens
$0.07 / $1.68
Model Inputs
Text
Model Outputs
Text
Tool Calling
Enabled

Compare With Similar Models