DeepSeek V3.2 Exp
by DeepSeek
DeepSeek-V3.2-Exp is an experimental large language model released by DeepSeek as an intermediate step between V3.1 and future architectures. It introduces DeepSeek Sparse Attention (DSA), a fine-grained sparse attention mechanism designed to improve training and inference efficiency in long-context scenarios while maintaining output quality. Users can control the reasoning behaviour with the `reasoning` `enabled` boolean. Benchmarking shows performance roughly on par with V3.1 across reasoning, coding, and agentic tool-use tasks, with minor tradeoffs and gains depending on the domain. This release focuses on validating architectural optimizations for extended context lengths rather than advancing raw task accuracy, making it primarily a research-oriented model for exploring efficient transformer designs.
Specifications
Technical details and pricing.
Benchmarks
10 benchmark scores from Artificial Analysis.
Composite Indices
Intelligence, Coding, Math
Standard Benchmarks
Academic and industry benchmarks
Frequently Asked Questions
What is DeepSeek V3.2 Exp good for?
Use DeepSeek V3.2 Exp for everyday tasks like writing, summarizing, brainstorming, and getting clear explanations.
How much does DeepSeek V3.2 Exp cost?
Pricing is based on usage. Current rates are $0.28/1M tokens for input and $0.42/1M tokens for output.
Can I try DeepSeek V3.2 Exp for free?
Yes. You can start a chat instantly and test the model before deciding on a plan.
Does DeepSeek V3.2 Exp support images or audio?
DeepSeek V3.2 Exp focuses on text-based tasks.
Similar Models
Other models you might want to explore.
Benchmarks and pricing are sourced from Artificial Analysis where available. OpenRouter specs are used as a fallback.