DeepSeek V4.1 Flash 1017 · Office #43,721

What it is

Calls itself “Flash model”.

• Calls itself: Flash model

“Flash model”

What it costs

Prices quoted: $0.30 in and $1.20 out per 1M tokens.

• Price: $0.30 in and $1.20 out per 1M tokens

“Peak price is $0.30 in and $1.20 out per 1M tokens.”

Who it sells to

Sells to developers. Made for coding agents.

• Sells to: developers
• Made for: coding agents

“It also adds a Search and Fetch API.”

“It has a 1M-token context and is tuned for coding agents like DeepSeek Harness.”

How it runs

A API and web app. Runs locally and works offline. Offers an API. AI is the core of the product.

• Kind: api · web
• Runs where: on the device, offline
• API: yes
• AI role: core to the product

“…served as deepseek-flash on the website, app and API.”

“Please refer to the inference folder for instructions on weight conversion and running inference locally.”

What it works with

Replaces DeepSeek-V4.1-Flash. Runs inside SGLang and vLLM. Names Claude Code and DeepSeek-V4.1-Flash-1017 on its pages.

• Replaces: DeepSeek-V4.1-Flash
• Runs inside: SGLang · vLLM
• Names: Claude Code · DeepSeek-V4.1-Flash-1017

“We use the Claude Code harness for Terminal-Bench 4.0”

“DeepSeek-V4.1-Flash-1017 is a new post-training update to DeepSeek’s Flash model”

What it offers as proof

Live, no waitlist. Instant signup. Publishes docs.

• Instant signup: yes

“Log In Sign Up”

“…encoding -> https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash-1017/blob/main/encoding/README.md”