The Orchestrator › Models › DeepSeek
DeepSeek: DeepSeek V3.2
DeepSeek-V3.2 is a large language model designed to harmonize high computational efficiency with strong reasoning and agentic tool-use performance. It introduces DeepSeek Sparse Attention (DSA), a fine-grained sparse attention mechanism...
Specifications
| Provider | DeepSeek |
| Model ID | deepseek/deepseek-v3.2 |
| Context window | 164K tokens (163,840) |
| Modalities | text → text |
| Tokenizer | DeepSeek |
| Router sources | OpenRouter, TrustedRouter, Vercel AI Gateway |
| Released | 2025-12-01 |
| Tracked since | 2026-11-14 |
| Popularity | #40 by usage · peak 2.8% spend share in an OpenRouter category |
| Reasoning | Supported |
Pricing (per million tokens)
| Variant | Source | API ID | Input | Output | Context |
|---|---|---|---|---|---|
| standard | OpenRouter | deepseek/deepseek-v3.2 | $0.26 | $0.38 | 164K |
| standard | TrustedRouter | deepseek/deepseek-v3.2 | $0.274 | $0.401 | 164K |
| standard | Vercel AI Gateway | deepseek/deepseek-v3.2 | $0.62 | $1.85 | 128K |
Recorded price history
| Date | Input / 1M | Output / 1M | Context |
|---|---|---|---|
| 2026-11-14 | $0.62 | $1.85 | 128K |
| 2026-11-14 | $0.274 | $0.401 | 164K |
| 2026-11-14 | $0.26 | $0.38 | 164K |
Capabilities
Agentic tool useStructured outputsControllable reasoningNo content storage (attested)Zero data retentionNo training on inputs (attested)
Trust flags come from TrustedRouter's per-provider attestations.
Supported parameters
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