DeepSeek V4 Flash

ServerlessDedicated
Hugging Face IDdeepseek-ai/DeepSeek-V4-FlashView on Hugging Face ↗

DeepSeek V4 Flash is the smaller model in DeepSeek's V4 preview series. It's a 284B-parameter mixture of experts with 13B active per token and a 1M-token context window.

Ways to run DeepSeek V4 Flash

Dedicated endpoint

Get a quote →

Your own GPUs, tuned for your traffic. We profile your workload, tune the deployment and put throughput and latency in your contract.

Serverless

API docs →

Pay per token on shared GPUs, with no capacity to plan or reserve.

Inside your network

Security →

Run a dedicated endpoint inside your own network, including air-gapped environments.

Pricing

Dedicated endpoint
Per GPU-hour
We size the GPUs to your traffic.
Serverless
$0.075 / $0.30
Per 1M tokens (input/output)

Use it with the agents we set up for you

Point any open-source agent at your endpoint, or pick from our catalog and we set it up as part of your contract.

Cline CodingOpenCode CodingAider Coding
GPT Researcher ResearchRAGFlow DocumentsNanobrowser Browser

DeepSeek V4 Flash FAQ

How much does DeepSeek V4 Flash cost on Saturn Cloud?

Serverless costs $0.075 per 1M input tokens and $0.30 per 1M output tokens. A dedicated endpoint is priced per GPU-hour for the GPUs it runs on, and we size the GPUs to your traffic.

What reasoning modes does DeepSeek V4 Flash have?

Three, called Non-think, Think High and Think Max. Think Max adds a special system prompt.

How does V4 Flash compare with V4 Pro?

V4 Pro is the larger model in the series, with 1.6T total and 49B active parameters and the same 1M context. DeepSeek says Flash-Max reaches reasoning comparable to V4 Pro when given a larger thinking budget.

What precision are DeepSeek V4 Flash's weights?

FP4 for the mixture-of-experts weights and FP8 for most other parameters.

What sampling settings does DeepSeek recommend?

Temperature 1.0 and top_p 1.0.