DeepSeek: DeepSeek V3.2 Exp

deepseek/deepseek-v3.2-exp

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...

Prompt $/1M

$0.270

Completion $/1M

$0.410

Context

163,840 tok

Best provider

21 tok/s · AtlasCloud

Specs

Model id: deepseek/deepseek-v3.2-exp
Display name: DeepSeek: DeepSeek V3.2 Exp
Modality: text->text
Tokenizer: DeepSeek
Instruct type: deepseek-v3.1
Context length: 163,840
Max completion: 65,536
Prompt $/1M: $0.270
Completion $/1M: $0.410
Free tier: no

Capabilities

Function callingno
JSON modeno
System promptyes
Visionno
Moderatedno

Benchmark scores

Each axis is one benchmark, normalized to the best model in the field.

No benchmark data for this model.

Speed by provider

Provider	tok/s	TTFT	Total latency
AtlasCloudfastest	21	1,466 ms	1,621 ms
Novita	20	1,406 ms	1,581 ms
SiliconFlow	15	1,677 ms	2,002 ms