DeepSeek · Model identity
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...
1pricing route
What published route metadata supports
CachingHugging FaceReasoningStructured outputTool use
Input modalities
Text
Supported API parameters
frequency_penaltyinclude_reasoninglogit_biaslogprobsmax_tokensmin_ppresence_penaltyreasoningrepetition_penaltyresponse_formatseedstopstructured_outputstemperaturetool_choicetoolstop_ktop_logprobstop_p
Only source-backed fields are filled
Release dateNot independently verified
Knowledge cutoffNot independently verified
Open weightsNot independently verified
OpenRouter catalog date2025-12-01
LLMPrice does not infer a release date, knowledge cutoff, or open-weight status from a model name, Hugging Face link, or pricing route.
Could this workload cost less?Compare DeepSeek V3.2 with cheaper models using the same workload.
Find cheaper alternatives →| Endpoint / route | Model ID | Input | Cached | Output | Context | Source |
|---|---|---|---|---|---|---|
| OpenRouterStandard | deepseek/deepseek-v3.2 | $0.269 | $0.1345 | $0.4 | 164K | OpenRouter API ↗ |
Hugging Face references
Presence of a Hugging Face ID is not treated as proof that model weights are open.
1 verified price change since Aug 26, 2026
0 route availability changes recorded. No earlier prices are inferred.
Input / 1M$0.269Aug 26, 2026 → Aug 28, 2026
Output / 1M$0.4Aug 26, 2026 → Aug 28, 2026
- Aug 28, 2026OpenRouter · Standard
Input $0.26 → $0.269 · Cached input $0.13 → $0.1345 · Output $0.38 → $0.4