Ahead of V4 Pro at Flash cost
DeepSeek reports V4.1 Flash beats V4 Pro on performance, cost, speed and total runtime in tests by multiple parties.
The DeepSeek V4.1 Flash API: DeepSeek's newest model, with image input and 1M context
deepseek-v4.1-flashfrom openai import OpenAIclient = OpenAI( base_url="https://api.zurelay.com/v1", api_key="YOUR_ZURELAY_KEY",)stream = client.chat.completions.create( model="deepseek-v4.1-flash", messages=[{"role": "user", "content": "Explain quicksort in two sentences."}], stream=True,)for chunk in stream: if chunk.choices: print(chunk.choices[0].delta.content or "", end="")Base URL https://api.zurelay.com/v1. Works with any OpenAI SDK.
Pricing
Pay as you go from prepaid credit, with no subscription and no minimum. Requests that fail are never billed.
| Rate | zurelay | DeepSeek list | You save |
|---|---|---|---|
Input per 1M tokens | $0.037 | 88%off | |
Output per 1M tokens | $0.15 | 88%off | |
Cached input per 1M tokens | $0.00074 | — | — |
One rate at every prompt length, up to the full 1M context window. Streaming, tool calls and structured outputs cost nothing extra.
Move the sliders to your monthly usage.
$283.80 a year
Overview
DeepSeek V4.1 Flash is DeepSeek's newest model and the one its own API now serves for Flash requests. It reads text and images, holds 1M tokens of context and runs in thinking or non-thinking mode. The zurelay DeepSeek V4.1 Flash API serves it through one OpenAI-compatible endpoint at 88% below DeepSeek's peak list price.
DeepSeek released V4.1 Flash on September 10, 2026 as the smallest model in its new architecture family, with native visual understanding. It is the model behind DeepSeek's deepseek-flash API name, and DeepSeek also routes its retired deepseek-v4-flash name to it. The DeepSeek V4.1 Flash API accepts text and images, returns text, and supports a 1M-token context with up to 384K output tokens.
Under the hood it is a 552B-parameter mixture-of-experts backbone with a new causal encoder-decoder design. It activates 8B parameters per token when reading input and 16B when generating output. With FP4 KV caching, the cache takes 890 bytes per token, about a quarter of V4 Flash. DeepSeek says this lowers the cost of running long-horizon agents at scale.
DeepSeek reports that tests by multiple parties put V4.1 Flash ahead of V4 Pro on performance, cost, speed and total runtime. Thinking mode is the default on DeepSeek's API, and you can switch it off when you want direct answers without a reasoning trace. The model supports tool calls and JSON output, and the weights are open under the MIT license.
On zurelay the model ID is deepseek-v4.1-flash, and DeepSeek's own name deepseek-flash works too, so code written for DeepSeek's API only needs a new base URL and key. You pay $0.037 per 1M input tokens and $0.15 per 1M output tokens at any hour. DeepSeek's own rates change between peak and off-peak hours.
Strengths
DeepSeek reports V4.1 Flash beats V4 Pro on performance, cost, speed and total runtime in tests by multiple parties.
A KV cache of 890 bytes per token, about a quarter of V4 Flash, keeps 1M-token prompts and long agent sessions affordable to serve.
Screenshots, charts and scanned documents go in alongside text, with no separate vision model.
zurelay accepts DeepSeek's deepseek-flash model name, so existing clients switch with a base URL and key change.
Use cases
Run agents that keep long tool histories in context. The small KV cache was designed for exactly this.
Use thinking mode for code generation, review and repository Q&A, and turn it off for quick completions.
Extract fields from scanned forms, charts and UI screenshots straight into JSON.
Summarize, classify and translate large volumes of text at a low per-token rate.
Get started
No waitlist and no new SDK. If your code already talks to OpenAI, it already talks to zurelay.
Sign up, add credit and create an API key. Set a monthly budget or a rate limit per key if you like.
Change the base URL. Everything else in your code stays the same.
https://api.zurelay.com/v1Send chat completions as usual. Streaming, tool calls and usage reporting work as you expect.
deepseek-v4.1-flashWorks with the tools you already use
Compare
$0.23 / $0.62 per 1M tokens
Pick DeepSeek V4 Pro 0813 when your prompts and evals are tuned on V4 Pro and you want that fixed checkpoint.
DeepSeek V4 Pro 0813 API$0.022 / $0.06 per 1M tokens
Pick DeepSeek V4 Flash 0731 to pin the previous-generation V4 Flash release for reproducible, text-only results.
DeepSeek V4 Flash 0731 API$0.045 / $0.18 per 1M tokens
Keep calling deepseek-v4-flash if your code already uses that name; like on DeepSeek's API, V4.1 Flash answers it.
DeepSeek V4 Flash APIFAQ
Anything else, write to sales@zurelay.com and an engineer will answer.
On zurelay, DeepSeek V4.1 Flash costs $0.037 per 1M input tokens and $0.15 per 1M output tokens, with cached input at $0.00074. DeepSeek's peak-hour list price is $0.30 input and $1.20 output, and it charges half that off-peak. zurelay uses one rate at all hours, 88% below DeepSeek's peak list price.
zurelay serves DeepSeek V4.1 Flash at 88% below DeepSeek's peak list price, with no peak-hour pricing to plan around. You pay as you go, and credit never expires.
Set the OpenAI SDK's base URL to https://api.zurelay.com/v1, use your zurelay API key and set model to "deepseek-v4.1-flash". If your code already uses DeepSeek's name deepseek-flash, zurelay accepts that too.
DeepSeek V4.1 Flash supports 1M tokens of context and up to 384K output tokens per response.
V4.1 Flash uses a new architecture: a 552B-parameter causal encoder-decoder, against V4 Flash's 284B-parameter model. It adds native image input and cuts the KV cache to about a quarter of V4 Flash's. DeepSeek retired V4 Flash on its own API when V4.1 Flash launched.
Agents, coding and long-context work where cost matters. DeepSeek designed it for faster inference, higher throughput and cheaper long-horizon agents, and reports it ahead of V4 Pro in tests by multiple parties.
Yes. It supports thinking and non-thinking modes, and thinking is the default on DeepSeek's own API. Use thinking for multi-step reasoning and coding, and non-thinking for quick, direct answers.
Streaming works with stream: true, as with the OpenAI API. You can set a requests-per-minute cap and a monthly budget on each zurelay API key.
Yes. Requests to deepseek-v4.1-flash run DeepSeek V4.1 Flash, the model DeepSeek serves as deepseek-flash. Responses are sampled, so wording varies between runs, as it does on DeepSeek's API. zurelay is independent and not affiliated with DeepSeek.
Create a key in seconds, point the SDK you already use at zurelay, and every request costs up to 85% less from the first token.