DeepSeek
Claude 3.7 Sonnet
Frontier
Claude 3.7 Sonnet (high output)
Frontier
Claude 3.7 Sonnet (thinking)
Frontier
Claude 4 Opus
Frontier
Claude 4 Opus (thinking)
Frontier
Claude 4 Sonnet
Frontier
Claude 4 Sonnet (thinking)
Frontier
Claude 4.1 Opus
Frontier
Claude 4.5 Sonnet
Frontier
Claude 4.5 Sonnet (thinking)
Frontier
DeepSeek V3.2 Exp
An experimental model hosted in the US introducing DeepSeek Sparse Attention for training and inference efficiency in long-context scenarios.
Model Specifications
Context window
163,840 tokens
Max output
32,000 tokens
Knowledge cutoff
September 2025
Multipart messages
No
Vision capabilities
No
Default Parameters
Temperature
0.3
Top P
1.0
Frequency penalty
0.0
Presence penalty
0.0
Top K
50
Supported Parameters
Temperature
SupportedControls randomness: Lower values make output more deterministic, higher values more creative.
Range: 0.0 - 2.0
Default: 0.3
More models from DeepSeek
DeepSeek V3.1 Terminus
A large hybrid reasoning model (671B parameters, 37B active) that supports both thinking and non-thinking modes. USA hosted, it improves tool use, code generation, and reasoning efficiency.
Context: 163,840
Output: 32,000
DeepSeek V3.1
DeepSeek's updated V3.1 model releasedhosted securely in the USA
Context: 160,000
Output: 30,000
DeepSeek V3 0324
DeepSeek's updated V3 model released on 03/24/2025 hosted securely in the USA
Context: 131,072
Output: 30,000