modelgrep
D

DeepSeek: DeepSeek V4 Flash

deepseek/deepseek-v4-flash

23rd smartest of 178Cheaper than 84% of paidReasoningToolsJSON
Use via OpenRouter ↗
Intelligence
46.0
23rd of 178
Design Elo
1278
3D
Speed
85
84th fastest
Latency
544ms
first token
Input price
$0.098
47th cheapest
Context
1.0M

How it compares

Smarter than87%
of all ranked models
Faster than72%
of all ranked models
Cheaper than84%
of all ranked models

Overview

DeepSeek V4 Flash is an efficiency-optimized Mixture-of-Experts model from DeepSeek with 284B total parameters and 13B activated parameters, supporting a 1M-token context window. It is designed for fast inference and...

Benchmarks

independent · via OpenRouter
Artificial Analysis85th percentile
Intelligence Index
46.0
Coding Index
39.8
Agentic Index
62.3
GPQA Diamond
87%
Humanity's Last Exam
28%
SciCode
42%
Tau²-Bench (agentic)
96%
Design Arena · Elo8,360 tournaments
3D
1278
Game Dev
1275
codecategories
1263
Website
1257
UI Component
1226
svg
1212
asciiart
1197
Data Viz
1178

Providers & pricing (18)

ProviderIn $/MOut $/MUptime
GMICloudfp8$0.098$0.19697.8%
Baidufp8$0.098$0.19799.9%
DeepInfrafp4$0.100$0.20099.9%
Cloudflare$0.100$0.20099.8%
DigitalOcean$0.105$0.21097.3%
SiliconFlowfp8$0.130$0.28099.7%
StreamLake$0.133$0.26699.3%
Alibaba$0.134$0.26899.9%
Morph$0.139$0.27899.6%
DeepSeekcache$0.140$0.28099.9%
Parasailfp8$0.140$0.28099.4%
AtlasCloudfp8$0.140$0.28099.7%
AkashMLfp8$0.140$0.28099.5%
Fireworks$0.140$0.28097.7%
Novitafp8$0.140$0.28099.9%
WandBfp8$0.140$0.28099.9%
Venice$0.170$0.35097.4%
Io Netfp8$0.204$0.39099.9%

Specifications

Context window1.0M
Max output
Knowledge cutoff
Input modalitiestext
Output modalitiestext
Prompt cachingSupported
Cache read price$0.020/M
ModeratedNo

DeepSeek V4 Flash FAQ

How much does DeepSeek V4 Flash cost?

DeepSeek V4 Flash costs $0.098 per million input tokens and $0.196 per million output tokens via OpenRouter, making it 47th cheapest of 298 paid models.

How smart is DeepSeek V4 Flash?

DeepSeek V4 Flash scores 46.0 on the Artificial Analysis Intelligence Index, ranking 23rd of 178 benchmarked models, with a GPQA Diamond score of 87%.

How fast is DeepSeek V4 Flash?

DeepSeek V4 Flash generates around 85 tokens per second with 544ms time-to-first-token (p50), the 84th fastest tracked model.

What is DeepSeek V4 Flash's context window?

DeepSeek V4 Flash supports a 1.0M-token context window. It accepts text input.

Compare head-to-head