This model is deprecated. We only continue performance benchmarking for the default 10k input token workload. Results for other workloads are historical and no longer updated.

DeepSeek has launched a newer model, DeepSeek V4 Flash 0731 (max). We suggest considering it instead.

For more information, see comparison of DeepSeek V4 Flash 0731 (max) to other models and API provider benchmarks for DeepSeek V4 Flash 0731 (max).

Open weights model

Released 2026年4月

DeepSeek V4 Flash (Reasoning, High Effort) API 服务商基准测试与分析

模型比较

本分析旨在帮助你根据使用场景,选择 DeepSeek V4 Flash (Reasoning, High Effort) 的最佳 API 服务商。

速度最快

#1
MakoraMakora
127.2 t/s
#2
NovitaNovita
117.5 t/s
#3
SiliconFlow (FP8)SiliconFlow (FP8)
117.2 t/s
#4
CoreWeaveCoreWeave
77.2 t/s
#5
Parasail (FP8)Parasail (FP8)
70.1 t/s

输出速度

共 7 家服务商

延迟最低

#1
MakoraMakora
10.54 s
#2
SiliconFlow (FP8)SiliconFlow (FP8)
12.05 s
#3
NovitaNovita
12.21 s
#4
CoreWeaveCoreWeave
17.62 s
#5
Parasail (FP8)Parasail (FP8)
19.03 s

首个答案 Token 延迟

共 7 家服务商

价格最低

#1
DeepInfra (FP4)DeepInfra (FP4)
$0.05
#2
GMIGMI
$0.05
#3
SiliconFlow (FP8)SiliconFlow (FP8)
$0.07
#4
NovitaNovita
$0.08
#5
MakoraMakora
$0.10

每 100 万 token 的混合价格

共 7 家服务商

可通过 7 家 API 服务商使用 DeepSeek V4 Flash (high),每家服务商的性能特征和价格各不相同。下方对比了各服务商的关键指标。

  • 输出速度最快的服务商是 Makora(127.2 t/s)、Novita(117.5 t/s)和SiliconFlow (FP8)(117.2 t/s)。 各服务商之间的速度差异显著,最快与最慢相差 183%。
  • 延迟方面,Makora(10.54 秒)、SiliconFlow (FP8)(12.05 秒)和Novita(12.21 秒) 的首个答案 Token 延迟最低。
  • 价格方面,DeepInfra (FP4)(0.05)、GMI(0.05)和SiliconFlow (FP8)(0.07) 每 100 万 token 的混合价格最低。 各服务商价格最多相差 2.2 倍。
  • Makora 兼具最高速度和最低延迟,性能表现最佳。若要优化成本,DeepInfra (FP4) 的价格最具竞争力。

亮点

Output tokens per second · Higher is better
Seconds · Lower is better
USD per 1M tokens (blended) · Lower is better

价格

Pricing: Cache Hit, Input, and Output

Price (USD per M Tokens) · Lower is better · 10,000 input tokens

Pricing: Blended Price

Blended at 7:2:1 (cache-input-output) · USD per 1M tokens (blended) · Lower is better

Pricing: Cache Discount

1 - (cache hit price / input price) · Higher is better

Output Speed vs. Price

Blended at 7:2:1 (cache-input-output) · Output speed: output tokens per second · USD per 1M tokens (blended)
Most attractive quadrant
Pareto line

速度

按输出速度(每秒 token 数)衡量

Output Speed: DeepSeek V4 Flash (Reasoning, High Effort)

Output speed: output tokens per second · 10,000 input tokens

Latency vs. Output Speed

Latency: seconds to first token received · Output speed: output tokens per second · 10,000 input tokens
Most attractive quadrant

延迟

按首 Token 延迟(秒)衡量

首个答案 Token 延迟

Seconds to first token received · Lower is better · 10,000 input tokens

端到端响应时间

Seconds to output 500 tokens, calculated based on time to first token, 'thinking' time for reasoning models, and output speed

端到端响应时间

Seconds to output 500 tokens, including reasoning model 'thinking' time · Lower is better · 10,000 input tokens

关键比较指标与 API 功能

CoreWeave 标志CoreWeave
1M
开放
$0.32
77
1.55
24.08
16.06
DeepInfra 标志DeepInfra
1.05M
开放
$0.10
48
0.89
37.11
25.81
GMI 标志GMI
1.05M
开放
--
--
--
--
--
Makora 标志Makora
1M
开放
$0.29
127
0.79
14.47
9.75
Parasail 标志Parasail
1.05M
开放
--
70
1.35
26.17
17.69
SiliconFlow 标志SiliconFlow
1.05M
开放
--
102
1.47
18.51
12.15
Novita 标志Novita
1.05M
开放
$0.12
118
1.66
16.46
10.55

常见问题

关于 DeepSeek V4 Flash (Reasoning, High Effort) 服务商的常见问题