---
id: "deepseek/deepseek-v3.2"
also: ["deepseek/deepseek-v3.2-20251201", "deepseek-ai/DeepSeek-V3.2", "hf/deepseek-ai/DeepSeek-V3.2", "requesty/deepseek/deepseek-v3.2", "vercel/deepseek/deepseek-v3.2"]
cheapest_list_usd_per_m: "0.2088 in, 0.3096 out, at gmicloud/fp8 on openrouter"
cheapest_effective_usd_per_m: "0.0921 in, 0.3093 out, at gmicloud/fp8 on openrouter"
context: 163840
max_output: 147456
input: ["text"]
output: ["text"]
released: "2025-12-01"
api: ["/api/v1/chat/completions", "/api/v1/completions", "/api/v1/messages", "/api/v1/responses"]
params: ["frequency_penalty", "include_reasoning", "logit_bias", "logprobs", "max_tokens", "min_p", "presence_penalty", "reasoning", "repetition_penalty", "response_format", "seed", "stop", "structured_outputs", "temperature", "tool_choice", "tools", "top_k", "top_logprobs", "top_p"]
endpoints: 25
gateways: "openrouter (15), vercel (5), hf (3), requesty (2)"
no_training: 20
trains: 0
retains_prompts: 7
coding: 44.2000
---
# DeepSeek: DeepSeek V3.2

DeepSeek‑V3.2 from DeepSeek harmonizes high computational efficiency with superior reasoning and agent performance. It builds on three main techniques: DeepSeek Sparse Attention for long‑context efficiency, a scalable reinforcement learning framework, and a large‑scale agentic task synthesis pipeline. This model excels at long-context reasoning and agentic tasks, efficiently handling extended inputs while maintaining strong accuracy. Its sparse attention design enables it to process complex, multi-step workflows without excessive compute costs. Overall, DeepSeek‑V3.2 targets long‑context reasoning, tool‑using agents, and efficient deployment in production environments.