DeepSeek V3.2

by DeepSeek · chat

DeepSeek V3.2 non-thinking mode. 671B MoE model (37B active) with 128K context. Excels at general chat, code generation, and agent tasks with integrated tool-use across 1,800+ environments.

Pricing

Input: $0.2484/M tokens · Output: .012/M tokens

Capabilities

Function Calling, JSON Mode, Streaming

Context: 128K tokens

Max output: 8K tokens

Routes: 6/6 healthy

Performance

TTFT: 5350ms · Latency: 14883ms · Throughput: 296.0 tok/s