DeepSeek: DeepSeek V4 Pro

deepseek/deepseek-v4-pro

DeepSeek V4 Pro is a large-scale Mixture-of-Experts model from DeepSeek with 1.6T total parameters and 49B activated parameters, supporting a 1M-token context window. It is designed for advanced reasoning, coding,...

Prompt $/1M

$0.435

Completion $/1M

$0.870

Context

1,048,576 tok

Best provider

101 tok/s · BaseTen

Specs

Model id: deepseek/deepseek-v4-pro
Display name: DeepSeek: DeepSeek V4 Pro
Modality: text->text
Tokenizer: DeepSeek
Instruct type: —
Context length: 1,048,576
Max completion: 384,000
Prompt $/1M: $0.435
Completion $/1M: $0.870
Free tier: no

Capabilities

Function callingno
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
BaseTenfastest	101	577 ms	845 ms
Artificial Analysis	74	977 ms	—
Fireworks	62	1,372 ms	2,764 ms
Alibaba	55	1,592 ms	1,817 ms
Baidu	52	727 ms	901 ms
Venice	48	1,370 ms	1,866 ms
GMICloud	47	2,746 ms	3,107 ms
AtlasCloud	46	2,225 ms	2,474 ms
Together	46	551 ms	690 ms
Novita	45	1,444 ms	1,710 ms
DeepInfra	44	728 ms	996 ms
DeepSeek	42	1,233 ms	1,451 ms
StreamLake	41	2,565 ms	3,279 ms
SiliconFlow	38	2,434 ms	3,040 ms
NextBit	38	2,044 ms	2,274 ms
DigitalOcean	31	878 ms	1,253 ms
Wafer	31	1,413 ms	4,081 ms
Parasail	28	614 ms	799 ms
Io Net	13	968 ms	968 ms
Ionstream	11	1,210 ms	1,498 ms
WandB	9	1,125 ms	1,542 ms