Back to Models

Deepseek v3.2

DeepSeek V3.2 is a 685B parameter MoE model leveraging DeepSeek Sparse Attention (DSA). It excels in complex mathematical reasoning and programming competitions, featuring integrated tool-use thinking modes.

Thinking Mode
Parameters
68500000000 B
Context
163,840 tokens
Released
Jan 12, 2025

Leaderboards

Average Score combining domain-specific Autobench scores; Higher is better

Performance vs. Industry Average

Intelligence

Deepseek v3.2 is of lower intelligence compared to average (2.9), with an intelligence score of 2.6.

Price

Deepseek v3.2 is cheaper compared to average ($0.75 per 1M Tokens) with a price of $0.06 per 1M Tokens.

Latency

Deepseek v3.2 has a higher average latency compared to average (44.25s), with an average latency of 53.60s.

P99 Latency

Deepseek v3.2 has a higher P99 latency compared to average (126.46s), taking 129.25s to receive the first token at P99 (TTFT).

Context Window

Deepseek v3.2 has a smaller context window than average (406k tokens), with a context window of 164k tokens.

Deepseek v3.2 - AutoBench