GPT Image 2.5 is live — OpenAI's newest image model, targeted edits that leave the rest of the frame alone

deepseek-v4.1-flash — 1M Context, 384K Output, Vision

deepseek-v4.1-flash is the DeepSeek-V4.1-Flash weights served on reAPI's OpenAI-compatible chat endpoint. deepseek-v4.1-flash reads a 1M-token context window and can write up to 384,000 tokens in a single response, thinks by default across three effort rungs, and takes images alongside text. Current rates for deepseek-v4.1-flash are in the pricing table on this page.

Highlights
1M context · 384,000 max output · vision · thinking on by default
Best for
Long documents, repository analysis, chart and screenshot reading
Input
Chat messages, text and image (OpenAI-compatible)
Output
Text, streaming optional
DeepSeek V4.1 Flash

Ask DeepSeek V4.1 Flash

Start with the prompt in the composer or choose an example.

Enter to send · Shift+Enter for a new line

Pricing

Credit-based — 1 credit = $0.001 USD. Pay only for completed generations.

First test cost
$0.00171.7 credits

One 1K-token prompt with a 1K-token answer at the cheapest tier.

Testing budget guide
Add credits
$10
≈ 5882 tests
$50
≈ 29411 tests
$100
≈ 58823 tests
Input
$0.400
Price · 1M tokens
Output
$1.300
Price · 1M tokens
Cache read
$0.015
Price · 1M tokens

What is deepseek-v4.1-flash?

deepseek-v4.1-flash is how reAPI serves DeepSeek's V4.1-Flash weights over an OpenAI-compatible /api/v1/chat/completions endpoint. DeepSeek's own Models and Pricing table gives the Flash line a 1M-token context window, a documented maximum of 384K output tokens, vision, JSON output and tool calls, and both thinking and non-thinking modes. Thinking is on by default at high effort. What makes deepseek-v4.1-flash unusual in its class is the output side: most fast tiers cap generation well below their context window, while deepseek-v4.1-flash can keep writing to 384,000 tokens, which is what a full translation pass or a file-by-file refactor actually needs. Billing is per token from your reAPI balance, with input, output and cache reads each carrying their own rate.

What you can build with this model

Real-world workflows and production use cases you can build and ship with this model.

deepseek-v4.1-flash processing a long document in a single pass

Whole-document and whole-repository passes

A 1M-token context window is only half of a long job; the other half is being allowed to write the answer out. deepseek-v4.1-flash pairs that window with a 384,000-token output ceiling, so a translated manual, a rewritten test suite or a clause-by-clause contract review can come back in one response instead of being chunked, stitched and reconciled. Because deepseek-v4.1-flash prices the same at any context length, a long prompt does not cross a threshold and change the rate mid-workload. Set max_tokens deliberately on these jobs — deepseek-v4.1-flash uses the endpoint default otherwise, not the ceiling.

Read the API docs
deepseek-v4.1-flash reading a chart and a screenshot

Charts, screenshots and scanned pages

deepseek-v4.1-flash takes images on the same endpoint and the same request shape as text. DeepSeek supports JPEG, PNG, GIF and WebP, and detects the format from the file's actual bytes rather than its name or declared MIME type, so a mislabelled upload still works. Send an image to deepseek-v4.1-flash inline as a base64 data URL, as a public http(s) link the model fetches for you, or as a Files API handle. That covers the two shapes this work normally arrives in: a local screenshot and a URL already sitting in your pipeline.

deepseek-v4.1-flash turning a scanned form into structured JSON

Structured extraction that stays structured

DeepSeek documents JSON output and tool calls as supported features of the Flash weights, so asking deepseek-v4.1-flash for machine-readable output is a contract rather than a prompt trick. Combined with vision, that makes deepseek-v4.1-flash a fit for turning scanned invoices, dashboards and forms into rows. One caveat worth knowing before you build the loop: when a request carries tools, the reasoning text deepseek-v4.1-flash returns must be echoed back on later turns; without tools it is ignored and can be dropped.

Why reAPI for DeepSeek V4.1 Flash

An output ceiling that matches the window

Fast tiers usually advertise a large context window and then cap generation an order of magnitude below it. DeepSeek documents a maximum of 384K output tokens for the Flash weights, so deepseek-v4.1-flash can return work at the scale it reads it. For translation, refactors and long structured extractions this is the difference between one request and a chunking pipeline you have to write, tune and debug yourself.

Vision on the same endpoint

deepseek-v4.1-flash accepts images in the standard OpenAI content-block format — no separate vision endpoint, no different base URL, no second model id to route to. Images can be inline base64, a public URL that deepseek-v4.1-flash downloads, or a Files API handle, and DeepSeek identifies the format from file content rather than the filename. Adding image understanding to an existing text integration is a change to the message body and nothing else.

One key across every model on the platform

The same reAPI key and the same balance that call deepseek-v4.1-flash also call the Claude, GPT, Gemini and Kimi families. Putting deepseek-v4.1-flash head to head with another model, or failing over when one provider has a bad hour, is a change to one model string rather than a second integration, a second invoice and a second set of credentials to rotate.

deepseek-v4.1-flash vs GPT-5.6 Luna

The real decision for a cost-sensitive tier with a very large context window. Both take text and images and both expose a reasoning dial, so the comparison is not about which is better overall — it is about output ceiling, how each one prices a long prompt, and how much control you get over thinking. The rate row is stated plainly: Luna is the cheaper of the two.

Capability
deepseek-v4.1-flash on reAPI
GPT-5.6 Luna
Context window
1M tokens, documented by DeepSeek for the Flash weights
1.05M tokens, documented by OpenAI
Max output tokens
384,000 — roughly three times the rival, and the reason a full pass can return in one response
128,000
Long-prompt pricing
One rate at any context length; a long prompt never crosses a pricing threshold
A second, higher tier applies once a prompt passes the documented threshold
Reasoning control
Three rungs — low, high, max — with high the default; other rungs are accepted and mapped onto these
Six rungs including none, so thinking can be switched off completely
Image input
Inline base64, a public URL fetched for you, or a Files API handle; format read from file content
Images accepted at their original dimensions
Rate per token
Higher than the rival on both input and output; the live figures are in the table above
Lower on both input and output

Comparison reflects behavior documented by DeepSeek and OpenAI at the time of writing. Rates are described as relationships rather than amounts; live numbers are in the pricing table above and on each model's own page.

Ship deepseek-v4.1-flash in three steps

  1. step 01

    Create a reAPI key

    Sign up and generate a key in your dashboard. New accounts start with credit already applied, which is enough to send real requests and read a real bill before you top up.

    Open
  2. step 02

    Point your base URL at reAPI

    Keep your OpenAI client. Swap the base URL to reapi.ai/api/v1 and the model string to deepseek-v4.1-flash — deepseek-v4.1-flash speaks the same Chat Completions contract, so nothing else in your call site changes.

    Open
  3. step 03

    Send messages and read the response

    POST your messages array and read choices[0].message. deepseek-v4.1-flash returns its reasoning in reasoning_content alongside content, and usage reports input, output and cached tokens separately.

    Open
docs/api/deepseek-v4-1-flash

API reference

Drop-in code and the full parameter table.

View full reference
curl https://reapi.ai/api/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-v4.1-flash",
    "messages": [
      { "role": "user", "content": "Hello" }
    ],
    "reasoning_effort": "high",
    "max_tokens": 8192,
    "stream": true
  }'

Parameter validation failed

Rejected synchronously with the failing field named. Check enum values, ranges, and URL fields — nothing is charged.

Authentication or balance issue

401 means a missing or invalid Bearer key; 402 means the reserve exceeds your balance. Manage keys and credits in the console.

Content or material rejected

Reference material and outputs pass automated moderation. Rejected tasks fail with a clear error and a full refund.

Task failed or timed out

A task that reaches a failed state is never charged — the reserve refunds automatically. Keep the task id and retry when ready.

Frequently asked questions

Common questions about this model.

deepseek-v4.1-flash is billed pay-as-you-go per token from your reAPI balance, with separate input, output and cache read rates and no subscription. The rate does not step up with context length, and reasoning tokens count as output. Current numbers are in the pricing table on this page.

Related models

Explore more models in the same category.

View all models
start building

Ready to ship?

Try it in the playground or grab an API key to integrate now.

Get API keyView API docs