# DeepSeek V4.1 Flash has landed on Pickaxe

**URL:** <https://community.pickaxe.co/t/deepseek-v4-1-flash-has-landed-on-pickaxe/8842>\
**Category:** Releases & Features\
**Created:** [September 15, 2026, 7:24pm UTC](https://community.pickaxe.co/t/deepseek-v4-1-flash-has-landed-on-pickaxe/8842 "2026-09-15T19:24:06Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![danny\_support](https://yyz2.discourse-cdn.com/flex004/user_avatar/community.pickaxe.co/danny_support/32/3492_2.png) [@danny\_support](https://community.pickaxe.co/u/danny_support)\
**Post date:** [September 15, 2026, 7:24pm UTC](https://community.pickaxe.co/t/deepseek-v4-1-flash-has-landed-on-pickaxe/8842/1 "2026-09-15T19:24:06Z")

</div>

> **deepseek?**
>

**DeepSeek V4.1 Flash is now available on Pickaxe!** It brings native image understanding, a one-million-token context window, and adjustable reasoning to DeepSeek’s Flash lineup, with support for Actions built in.

**Images and text, together**

V4.1 Flash can work with images alongside your written instructions. Screenshots, charts, diagrams, and photographed documents can all become part of the conversation.

> **[deepseek-ai/DeepSeek-V4.1-Flash · Hugging Face](https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash)**
>
> We’re on a journey to advance and democratize artificial intelligence through open source and open science.

**A million tokens of working room**

The model supports a context window of up to **one million tokens** , giving it room for substantial reference material and extended conversations. Context is the material available to the model during a request, including instructions, supplied information, and conversation history.

It also supports up to **384,000 output tokens** , providing capacity for lengthy responses and reasoning. These are the model’s maximum capacities; individual requests use the space they need.

> **[Models & Pricing | DeepSeek API Docs](https://api-docs.deepseek.com/quick_start/pricing/)**
>
> The prices listed below are in units of per 1M tokens. A token, the smallest unit of text that the model recognizes, can be a word, a number, or even a punctuation mark. We will bill based on the total number of input and output tokens by the model.

**The engineering behind Flash**

One interesting detail is how the model divides its work. Its backbone contains **552 billion parameters** , but it activates **8 billion per token when processing input** and **16 billion when generating output**.

This comes from a combination of a mixture-of-experts design and DeepSeek’s new encoder–decoder architecture. The goal is to keep computation efficient, particularly for tasks that involve reading a large amount of information before producing an answer.

> **[Introducing DeepSeek-V4.1-Flash: smarter, faster, more efficient.](https://www.deepseek.com/en/news/deepseek-v4-1-flash/)**
>
> Introducing the smallest model in our new architecture family, with native visual understanding. Designed for greater capability, faster inference, and higher throughput.

DeepSeek has also released the model weights under the **MIT license**. [Model release and license](https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash)

**A look at the benchmarks**

DeepSeek reports the following results in its [model card](https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash):

| Benchmark | What it tests | Reported result |
| --- | --- | --- |
| Chartography, with tools | Understanding and answering questions about professional charts | **78.9%** |
| GPQA Diamond | Difficult graduate-level science questions | **90.9%** |
| DeepSWE v1.1 | Completing substantial software-engineering tasks | **74.2% resolved** |

_These are provider-reported results at maximum reasoning effort, using the evaluation setups described in DeepSeek’s model card._

**Try it on Pickaxe**

DeepSeek V4.1 Flash is in the model selector now, with **image uploads, reasoning, and Actions** supported. You can find current pricing and the full specifications on our

> **[DeepSeek V4.1 Flash | AI Models | Pickaxe](https://pickaxe.co/models/deepseek-v4-1-flash)**
>
> DeepSeek's fast, cost-efficient multimodal model for reasoning, coding, image understanding, tool use, and long-context agent workflows.

Give it a try and let us know how it’s working for you!