DeepSeek: DeepSeek V3.2

deepseek/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...

Prompt $/1M

$0.269

Completion $/1M

$0.400

Context

163,840 tok

Best provider

63 tok/s · DeepSeek

Specs

Model id: deepseek/deepseek-v3.2
Display name: DeepSeek: DeepSeek V3.2
Modality: text->text
Tokenizer: DeepSeek
Instruct type: —
Context length: 163,840
Max completion: 65,536
Prompt $/1M: $0.269
Completion $/1M: $0.400
Free tier: no

Capabilities

Function callingyes
JSON modeno
System promptno
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
DeepSeekfastest	63	940 ms	1,176 ms
GMICloudfastest	46	2,242 ms	2,381 ms
Google	38	1,437 ms	1,728 ms
Alibaba	29	1,400 ms	1,744 ms
Friendli	28	2,567 ms	3,564 ms
DigitalOcean	24	794 ms	1,098 ms
StreamLake	22	1,368 ms	1,558 ms
Baidu	21	1,062 ms	1,184 ms
Novita	21	1,195 ms	1,404 ms
AtlasCloud	20	1,270 ms	1,494 ms
SambaNova	18	3,395 ms	3,591 ms
SiliconFlow	17	1,818 ms	2,346 ms
Parasail	12	938 ms	1,579 ms
AkashML	11	2,615 ms	3,001 ms
DeepInfra	11	967 ms	1,313 ms
Phala	8	4,200 ms	4,664 ms
Venice	8	1,201 ms	1,635 ms
Chutes	5	2,692 ms	3,338 ms
Artificial Analysis	0	0 ms	—