DeepSeek-V3.2-Speciale is a high-compute version of DeepSeek-V3.2, designed for maximum reasoning and agentic performance. It builds on DeepSeek Sparse Attention (DSA) for efficient long-context handling and uses scaled reinforcement learning post-training to surpass the base model. Evaluations show that Speciale outperforms GPT-5 on complex reasoning tasks and matches Gemini-3.0-Pro in proficiency, while maintaining strong coding and tool-use reliability. Like V3.2, it uses a large-scale agentic task synthesis pipeline to enhance compliance and generalization in interactive environments.
Retired on Jan 27, 2026: requests to this model ID now return an error. These models are live and work with the same API key.
DeepSeek-V3.2-Speciale is a high-compute version of DeepSeek-V3.2, designed for maximum reasoning and agentic performance. It builds on DeepSeek Sparse Attention (DSA) for efficient long-context handling and uses scaled reinforcement learning post-training to surpass the base model.
No. Deepseek v3.2 Speciale was retired on NagaAI on Jan 27, 2026, and requests to it now return an error. Live models you can use instead with the same API key: DeepSeek V4.1 Flash 1017, DeepSeek V4.1 Flash 0910 and DeepSeek V4 Flash Vision Exp.
It takes text as input and returns text.
