Summary
The model deepseek/deepseek-v3.2-speciale is available on OpenRouter but is missing from the model catalog. The catalog already has deepseek/deepseek-v3.2 and deepseek/deepseek-v3.2-exp but not the Speciale variant.
DeepSeek-V3.2-Speciale is a high-compute variant of V3.2 optimized for maximum reasoning and agentic performance. It uses scaled post-training reinforcement learning and a large-scale agentic task synthesis pipeline. It is architecturally the same MoE (671B total, 37B active) but trained differently for deeper chain-of-thought and tool-calling scenarios.
Verification
Suggested spec
{
"deepseek/deepseek-v3.2-speciale": {
"format": "openai",
"flavor": "chat",
"reasoning": true,
"max_input_tokens": 131072,
"available_providers": ["openrouter"]
}
}
Note on pricing: OpenRouter search results suggest ~$0.21/M input and ~$0.31/M output, but these figures should be verified against the live OpenRouter API before committing. Azure AI Foundry lists different pricing ($0.58/$1.68) for its own deployment.
Summary
The model
deepseek/deepseek-v3.2-specialeis available on OpenRouter but is missing from the model catalog. The catalog already hasdeepseek/deepseek-v3.2anddeepseek/deepseek-v3.2-expbut not the Speciale variant.DeepSeek-V3.2-Speciale is a high-compute variant of V3.2 optimized for maximum reasoning and agentic performance. It uses scaled post-training reinforcement learning and a large-scale agentic task synthesis pipeline. It is architecturally the same MoE (671B total, 37B active) but trained differently for deeper chain-of-thought and tool-calling scenarios.
Verification
deepseek/deepseek-v3.2-specialematches established OpenRouter naming (provider/model-variant)Suggested spec
{ "deepseek/deepseek-v3.2-speciale": { "format": "openai", "flavor": "chat", "reasoning": true, "max_input_tokens": 131072, "available_providers": ["openrouter"] } }