DeepSeek V4.1 Flash (US)

Optional online service

DeepSeek · Oct 2026 · Hosted in the United States by Fireworks AI

DeepSeek's fast, low-cost model, served from US data centers by Fireworks - quick everyday answers, with images.

Context
1M tokens
Input
$0.58/M
Output
$2.34/M
Vision
Reads images
chatcoding

Performance

Our questions
36 of 36 right
Time for an answer
1 second

Number 1 of 34 by time (1 is the quickest)

Cost of all 36 answers
2 cents

Number 7 of 34 by cost (1 is the cheapest)

We ask every online model the same 36 questions, each with one answer that can be checked: sums in several steps, logic, counting, reading code and a table, and following an instruction to the letter. 31 of the 34 models asked got every one right, so the difference is mostly in how long a model takes and what it costs to get there. The cost counts the model's thinking, which a price per token does not show: a model with a low price can write many times more. Asked on November 16, 2026, with each provider's own settings.

How Your Own AI grades it

  • General8 / 10
  • Coding8 / 10
  • Reasoning8 / 10
  • Maths8 / 10
  • Images7 / 10
  • Using tools7 / 9

These grades are what the app's automatic routing goes by when it picks an online model for a question. They are our own judgment from the maker's published results and from using the model, kept in check by the questions above.

Pricing

Input / M tokensOutput / M tokensCached input / MPer-request fee
$0.58$2.34$0.01None

An optional paid service: sign in with your Flowsta Identity to use it. Billed by actual tokens used, against your allowance and metered rates - the app shows the cost of every conversation as you go.

Privacy

Online models see only what you choose to send. You sign in with your Flowsta Identity - created by you, kept in your own Flowsta Vault, 24 words as the key - so there is no provider account behind your requests: they reach the provider with no identity attached. Your Own AI keeps your conversation history, memory, and documents on your device, and attachments only go online with your explicit consent - the app asks. License: MIT (open weights, served via API) (terms).

Frequently asked questions

How do I use DeepSeek V4.1 Flash (US) without a DeepSeek account?
Through Your Own AI: online models are an optional paid service you sign in to with your Flowsta Identity - one identity, created and kept by you, covers every model in the catalog. You pay per use; no provider subscriptions, no separate keys.
What does DeepSeek V4.1 Flash (US) cost?
$0.58 per million input tokens and $2.34 per million output tokens, with cached input at $0.01. The cost explorer on this page turns that into a monthly figure for your usage.
How good is DeepSeek V4.1 Flash (US)?
On Your Own AI's own set of 36 questions with one checkable answer, DeepSeek V4.1 Flash (US) got 36 right, took about 1 second an answer and cost 2 cents for the whole set at our price, thinking included. Of the 34 online models asked, it was number 7 by cost and number 1 by time, where 1 is the least.
How much context does DeepSeek V4.1 Flash (US) support?
1M tokens of context.
Is my data used to train DeepSeek V4.1 Flash (US)?
You sign in with a Flowsta Identity you hold yourself, so there is no provider profile to train against: requests reach the provider with no account or identifier attached, and Your Own AI keeps your history on your device. Attachments only go online with your explicit consent.
Is DeepSeek V4.1 Flash (US) open source?
The weights are published under the MIT license - this catalog serves it as a hosted API for speed, and you can run open models like it locally with Your Own AI's offline catalog.

Use DeepSeek V4.1 Flash (US) the private way

Download the app, then sign in with a Flowsta Identity - yours to create, yours to keep - to add the online service. Pay per use, everything local stays local. No one in control but you.