# DeepSeek: DeepSeek V4.1 Flash

Provider: deepseek  
Category: llm  
Model ID: `deepseek/deepseek-v4.1-flash`

DeepSeek V4.1 Flash is a sparse mixture-of-experts model from DeepSeek, and the first built on the company's Causal Encoder-Decoder (CED) architecture. It activates 8B parameters on input and 16B on...

## Specs

- Context length: 1048576 tokens
- Max output: 943718 tokens
- Modalities: text, image
- Released: 2026-09-10

## Pricing

- Input: $0.05 per million tokens
- Output: $0.12 per million tokens

## Providers

- **deepseek** — ctx 1048576, input $0.30/M, output $1.20/M

## Verdict

DeepSeek V4.1 Flash delivers strong general-purpose performance at a price point that undercuts GPT-4o and Claude by 40-60%. The million-token context window handles book-length documents without chunking, and vision support covers screenshots and diagrams. Speed trades off against top-tier reasoning — it won't match o1 or Claude Sonnet 4.5 on complex multi-step problems, but for most production workloads where cost and throughput matter, this is the model to reach for.

---
Last verified: 2026-10-19T02:30:34.239Z  
Canonical URL: https://switchy.build/directory/models/deepseek-v4-1-flash